With anonymous bulk data extraction ending, patent practitioners face a critical operational deadline. To maintain operational efficiency and technical competitiveness, patent professionals should actively secure an Application Programming Interface (API) key.

As artificial intelligence integration becomes standard practice for legal departments, possessing direct, programmatic access to patent data is a foundational requirement for modern prosecution and litigation strategies.

The Necessity of Programmatic Access for Artificial Intelligence

The patent system is experiencing a rapid technological transition. Forward-thinking law firms and corporate legal departments are actively developing and deploying internal artificial intelligence tools. These applications perform sophisticated tasks, ranging from predicting examiner behavior and analyzing complex prior art to automating the drafting of office action responses.

Legal technology companies release new generative models specifically trained on patent claims and specifications frequently. To utilize these external platforms securely or build proprietary models internally, practitioners require an uninterrupted flow of accurate, official data.

These internal and commercial platforms require massive streams of structured data to function correctly. Previously, developers could extract data anonymously or download massive XML files from legacy systems like the Patent Examination Data System or the Bulk Data Storage System. The agency has systematically discontinued those methods. The official FAQ explicitly states that “data will no longer be available in XML files; it will be available in JSON format” via the API (FAQ, ¶ 62).

To feed custom internal AI systems, practitioners need authorized, programmatic access. Securing an API key allows a firm’s technical staff, or its approved external software vendors, to connect directly to the agency’s databases. This authenticated connection bypasses the manual web interface, permitting the high-volume, automated data retrieval necessary for advanced analytical tools. The USPTO clarifies the new technical reality, stating that “ODP’s APIs provide developers with programmatic access and require an API key” (FAQ, ¶ 69).

Eligibility Parameters and Administrative Restrictions

Who is permitted to obtain these digital credentials? The agency’s policy enforces strict, individual accountability. The administration mandates that “[a]n API key is by person, not organization, and verified through a personal ID.me account” (FAQ, ¶ 81).

Law firms cannot acquire a single master key or a corporate license for an entire department. Every attorney, patent agent, paralegal, or data scientist requiring programmatic access must obtain their own credentials.

The agency firmly rejects requests for organizational exemptions, stating they “cannot make exceptions for organizations as the personal identifier verification process enhances our security for online services” (FAQ, ¶ 81). International users and foreign associates are eligible to participate but face a different verification track.

The agency specifies that international stakeholders “must verify your identity on a video call if you live outside the United States” (FAQ, ¶ 76).

The Credential Acquisition Process

The procedure to acquire the key involves several distinct administrative steps:

1. Establish a USPTO.gov Account: The individual must first create a standard USPTO.gov account.

2. Link a Verified ID.me Profile: The agency instructs that users must “sign in to the MyODP page with your USPTO.gov account, validate, and link your ID.me account” (FAQ, ¶ 71). This integration operates as a “one-time registration” (FAQ, ¶ 72).

3. Generate and Manage the Key: Following successful identity verification, the user can generate the API key directly from their personal MyODP dashboard. The MyODP page also functions as the central administrative hub for credential management. Users have the ability to manually deactivate or reactivate their API key directly from this dashboard to block access when necessary. This becomes a critical administrative step when personnel changes occur, as an employee leaving an organization can and should deactivate their API key on their personal MyODP page to sever programmatic access (FAQ).

4. Integrate the Key: Developers then integrate the key into their software requests. The agency provides a sample format for the HTTP header requiring the key (FAQ, ¶ 77).

The agency imposes this technical requirement “[t]o manage the load on the system and ensure access to USPTO data is available and fair to everyone” (FAQ, ¶ 70).

Before adopting an API-centric workflow, assessing the strategic benefits, logistical challenges, and operational risks provides necessary context for firm management.

Benefits

Securing an API key provides distinct advantages for the modern patent practice. Possessing this credential allows patent professionals to build or utilize advanced artificial intelligence platforms that require real-time, structured patent data. Rather than relying on third-party public portals that may experience latency or unexpected closures, practitioners maintain a direct pipeline to the official record.

This access empowers firms to automate the retrieval of file wrappers, monitor competitor portfolios efficiently, and conduct comprehensive statistical analyses of Patent Trial and Appeal Board proceedings. Firms controlling their own API access reduce their reliance on expensive, external commercial databases.

Direct API connections facilitate customized internal dashboards tailored specifically to the needs of the firm’s clients, providing real-time alerts regarding patent application status changes or newly cited prior art references.

Challenges

The transition to programmatic access introduces immediate administrative burdens. Law firm management must coordinate the ID.me verification process for numerous employees, a task that often encounters technical delays.

Users face strict operational boundaries once the key is acquired. The agency enforces API rate limits, warning that if a user submits “more than one request at a time or exceed weekly quota, you will get the error message: ‘HTTP 429 Too Many Requests’” (FAQ, ¶ 66).

The penalty for exceeding the download quota disrupts workflows severely; users “will need to wait seven days from the time you first started using the API key to try again” (FAQ, ¶ 67).

Risks

Moving toward individualized, API-driven data access carries operational vulnerabilities. Since the keys are tied to individual practitioners rather than the organization, a departing employee creates an immediate data vacuum.

The FAQ notes that “[a]n ID.me account cannot be reassigned as an API key is assigned by person, not organization” (FAQ, ¶ 79). Firms risk losing access to critical data pipelines if the sole technical key holder resigns or becomes unavailable.

Relying on a third-party identity verification service introduces external dependencies; if ID.me experiences outages, patent professionals cannot generate new keys or authenticate existing systems.

Integrating external AI platforms via personal API keys raises data security considerations, demanding strict oversight of how vendors handle the retrieved government records to prevent accidental breaches of client confidentiality.

Conclusion

The impending closure of anonymous Open Data Portal access forces the intellectual property community to adapt its data retrieval methods. Securing an API key is the logical response to this regulatory shift.

Patent professionals planning to leverage artificial intelligence for prosecution or litigation analytics require reliable, programmatic access to official records. By understanding the individual verification requirements and completing the registration process, practitioners maintain a competitive technical advantage.

The procedural hurdles of identity verification represent a necessary adaptation to a legal environment increasingly dependent on structured, machine-readable data.

Disclaimer: This is provided for informational purposes only and does not constitute legal or financial advice. To the extent there are any opinions in this article, they are the author’s alone and do not represent the beliefs of his firm or clients. The strategies expressed are purely speculation based on publicly available information. The information expressed is subject to change at any time and should be checked for completeness, accuracy and current applicability. For advice, consult a suitably licensed attorney and/or patent professional.

For patent practitioners and intellectual property professionals, this represents another systemic shift toward verified, authenticated access to government records.

The Open Data Portal has functioned as the unified data platform for the agency since February 2025, providing publicly accessible patent and trademark datasets.

Soon, the front door to this repository will require a digital key.

Closing Another Door to Anonymous Access

Observers of patent office policy may recognize a distinct pattern regarding data access. In September 2025, the agency eliminated guest access to Patent Center. That regulatory update required users to authenticate via ID.me or a paper verification form to view prosecution histories.

The policy disturbed the workflows for many firms and IP stakeholders, forcing support staff, in-house counsel, and foreign associates to adopt new verification procedures. Following that restriction, practitioners heavily utilized the Open Data Portal as an alternative method to view public file histories without complex login procedures.

A September 2025 blog post highlighted Global Dossier and the Open Data Portal web interface as the two remaining avenues for anonymous file wrappers. The impending June 2026 update removes the ODP from that short list of anonymous channels.

The agency previously implemented interface modifications to block automated actions. In April 2026, the USPTO restricted sponsored support staff from inserting practitioner signatures on the Web Application Data Sheet.

That procedural barrier appeared specifically formulated to throttle the volume of machine-generated filings. The current ODP policy aligns directly with these prior actions. Patent administrators are actively creating an environment where the agency can monitor and control the flow of intellectual property data into and out of its servers.

Targeting Automated Scrapers and AI Data Mining

The explicit targeting of unregistered traffic suggests the USPTO aims to minimize misuse of the portal by automated systems and scrapers. Large language models, such as Claude and other generative artificial intelligence systems, require massive datasets for training and operation.

Patent records feature highly technical, standardized language, making them highly desirable for machine learning developers. Unrestricted access permits automated programs to extract millions of documents, which apparently can stress government infrastructure.

The USPTO notes that mandatory sign-ins will “[b]lock harmful automated bot traffic that can slow or disrupt the site” and “[r]educe infrastructure costs” (FAQ, ¶ 14). By placing the data behind a registration wall (or API key), the agency can identify and potentially restrict users extracting bulk data for AI training without authorization.

The FAQ explicitly warns developers about API rate limits, noting that users exceeding their weekly download quota “will receive an HTTP 429 error” and must “wait seven days from the time you first started using the API key to try again” (FAQ, ¶ 73). These mechanical throttles demonstrate a clear intent to slow down machine-speed extraction.

New Profiles Coming

The agency explicitly connects this registration mandate to gathering more granular data on its users, noting the requirement “will allow the agency to improve the customer profile and future offerings” (Announcement, ¶ 3).

The administration plans to expand this data collection soon, stating they expect “to send a request for more profile information late this summer to improve existing customer profiles” (Announcement, ¶ 4). Patent practitioners and IP stakeholders might view this data gathering cautiously.

Detailed profiling establishes a framework for the granular monitoring of research habits. Sorting users into specific categories—such as law firm, academic institution, or commercial data broker—could provides the administrative architecture necessary to implement tiered access, restrict specific demographic groups, or introduce commercial usage fees at a later date.

Technical Shifts: The End of XML and Legacy Platforms

Practitioners adjusting to the June 18 deadline should evaluate their data extraction methods. The Developer Hub (DH) will retire entirely on May 29, 2026 (FAQ).

All public data from the DH has migrated to the ODP. Users currently accessing Application Programming Interfaces (APIs) will continue using their accounts, but obtaining a new API key requires strict ID.me verification. The FAQ specifies that “[i]n order to get access to an API you must get your USPTO.gov account verified with ID.me to receive an API key” (FAQ, ¶ 16).

The consolidation of legacy platforms into the ODP makes compliance unavoidable. The portal now houses data previously distributed across multiple systems. The FAQ lists the inclusion of Patent Trial and Appeal Board (PTAB) proceeding data, Patent File Wrapper (PFW) datasets previously found in the Patent Examination Data System (PEDS), Final Petitions Decisions (FPD), and the Bulk Data Storage System (BDSS) (FAQ, ¶ 6). The Bulk Data Storage System webpage “has been retired,” forcing users to access bulk files through the ODP’s directory feature (FAQ, ¶ 35).

The transition away from legacy systems shows a shift in data handling. Many patent professionals relied on the Patent Examination Data System to download the entire USPTO dataset in flat, static XML files. The agency has discontinued this practice. The FAQ confirms that data will no longer be available in XML files and specifies that users must now access this information programmatically in JSON format via the API (FAQ, ¶ 62). This technical shift requires developers to rewrite their data ingestion pipelines entirely. Software tools built to parse XML bulk downloads will break unless updated to handle the new JSON API endpoints.

In addition to formatting changes, the search parameters have been modified. The new interface allows users to conduct Boolean queries across several attributes from one search bar and customize the data attributes they wish to view (FAQ, ¶ 40).

For practitioners accustomed to the old systems, these changes require dedicated retraining time. The agency affirms that an API key is assigned “by person, not organization” and rejects requests for company-wide API keys (FAQ, ¶ 81). This policy forces every individual developer or data scientist within a legal technology company to undergo personal identity verification.

Benefits

Requiring registration for ODP access provides measurable advantages for system stability. Unregistered bot traffic consumes significant server bandwidth. By forcing users to log in, the USPTO can allocate resources more effectively, maintaining a platform that remains safe, fast, and reliable.

Verified accounts allow administrators to track usage patterns and prevent malicious actors from degrading site performance. For regular practitioners, this may result in faster load times and fewer system outages during critical research tasks.

The consolidation of data into a single, modernized platform improves the quality of the information extracted. The USPTO confirms that the data is preprocessed and analyzed, confirming schema consistency (FAQ, ¶ 46). Users extracting Patent File Wrapper data can now download specific JSON responses rather than downloading massive, static XML files to find a single record.

Challenges

The updated policy introduces administrative friction for IP professionals and support staff. Law firms must verify all personnel requiring public patent data have active, registered USPTO.gov accounts. Foreign associates and independent researchers face additional steps to view formerly public records. Consolidating datasets into the ODP means users cannot easily circumvent the login requirement by visiting legacy platforms like the Bulk Data Storage System. Every inquiry requires authentication.

The prohibition on shared organizational API keys creates logistical hurdles for corporate legal departments and patent analytics companies. If a developer leaves an organization, their ID.me-linked API key cannot be reassigned (FAQ, ¶ 79). Firms and companies must direct new employees to create and verify new personal API keys, potentially slowing down onboarding processes for technical staff.

Risks

Restricting access to public data carries systemic vulnerabilities. The patent system operates on the principle of public disclosure in exchange for exclusive rights. Placing the entire repository of bulk patent data behind a registration requirement may create a barrier for independent inventors, journalists, and academic researchers.

The agency states that ODP provides access to data “made available to the public as set forth in 37 CFR 1.11 and 1.14” (FAQ, ¶ 5). Yet, adding mandatory registration limits immediate transparency.

Developers relying on ODP APIs face the stringent ID.me verification process. If the verification systems experience downtime, third-party commercial patent software reliant on USPTO APIs could suffer catastrophic failures, leaving patent attorneys without necessary analytics tools. The reliance on a third-party identity verification service introduces a point of failure outside the direct control of the patent office.

The agency risks penalizing legitimate automated research in its pursuit of harmful bot traffic. Academic institutions utilizing algorithms to study patent trends may trigger the HTTP 429 error rate limits, delaying valuable economic and legal research.

By treating all automated extraction with suspicion, the agency may slow the development of legitimate, helpful AI tools built to assist practitioners.

Conclusion

The USPTO decision to mandate registration for the Open Data Portal marks a definitive end to anonymous bulk data extraction. The administration is actively constructing digital barriers to manage the heavy demands of automated AI scrapers like new versions Claude and ChatGPT.

Patent practitioners, data developers, and IP stakeholders must adapt their research workflows to accommodate these new verification standards prior to the June 18 deadline. The agency continues to prioritize system security and resource management over unfettered, anonymous public access.

For routine file wrapper inspection, Global Dossier stands as the final public portal offering anonymous access. The Global Dossier system operates as a collaborative IP5 initiative rather than an exclusive USPTO property. The platform escapes the immediate reach of the June 2026 login mandate.

Patent professionals can continue using the interface for single-document retrievals of post-2003 applications. Placing heavy reliance on Global Dossier carries strategic risks. The system lacks bulk download capabilities and offers no programmable API for automated extraction.

If data scrapers redirect their traffic toward Global Dossier following the ODP registration deadline, the participating international offices might implement parallel authentication barriers to protect their shared infrastructure.

The patent community should prepare for a scenario where easy, anonymous access to official patent records ceases to exist—and eventually requires a subscription.

Disclaimer: This is provided for informational purposes only and does not constitute legal or financial advice. To the extent there are any opinions in this article, they are the author’s alone and do not represent the beliefs of his firm or clients. The strategies expressed are purely speculation based on publicly available information. The information expressed is subject to change at any time and should be checked for completeness, accuracy and current applicability. For advice, consult a suitably licensed attorney and/or patent professional.

In United States v. Heppner, Judge Jed S. Rakoff of the Southern District of New York issued a memorandum opinion that will likely serve as the foundational text for AI privilege disputes.

By ruling that a criminal defendant’s communications with Anthropic’s “Claude” were protected by neither attorney-client privilege nor the work product doctrine, the court has put practitioners on notice.

Heppner Memorandum 02-17-2026

2.56MB ∙ PDF file

Download

Download

The opinion is not just a ruling on facts; it is a roadmap of what not to do. For litigators and corporate counsel, here are the essential takeaways from the Heppner decision.

1. The “Behest” Requirement is Non-Negotiable

The most significant finding for practitioners regarding the work product doctrine lies in the concept of “direction.” The defendant, Heppner, lost his work product claim largely because he acted independently.

The court emphasized that the documents “were prepared by the defendant on his own volition” (p. 10). Crucially, the court cited defense counsel’s own admission that they “did not direct [Heppner] to run Claude searches” (p. 10). Because of this lack of direction, the materials could not be said to be “prepared by or at the behest of counsel” (p. 9).

The Takeaway: To preserve a work product claim over AI-generated material, the attorney must be the architect of the interaction. Passive receipt of AI outputs generated by a client is insufficient. Practitioners should consider explicitly memorializing instructions to clients or consultants (and updating it if/when things change).

If AI is to be used, the engagement letter or a specific directive should state: “Client is directed to use [Specific Enterprise Tool] to organize these specific facts to assist Counsel in rendering legal advice.” Without that explicit “behest,” the court views the AI output as a mere client diary entry—fully discoverable.

2. The Terms of Service constitute a “Third-Party Waiver”

Practitioners often focus on the content of the communication, but Judge Rakoff focused on the contract governing the tool. The court dismantled the privilege claim by reading the platform’s Privacy Policy.

The court noted that the policy allowed the vendor to “collects data on both users’ ‘inputs’ and Claude’s ‘outputs’” and to use that data to “’train’ Claude” (p. 6). Most damningly, the policy reserved the right to “disclose personal data to third parties” (p. 6).

Based on these standard terms, the court held:

“Heppner could have had no ‘reasonable expectation of confidentiality in his communications’ with Claude” (p. 7).

The Takeaway: Standard, public-facing AI tools (the free or “pro” versions of ChatGPT, Claude, Gemini, etc.) are effectively crowded elevators. Practitioners must audit the tools their firms and clients use. If the Terms of Service (ToS) allow for model training or third-party review, privilege is likely waived the moment the “Enter” key is pressed. Practitioners must migrate to “Enterprise” or “API” tiers where data retention is contractually zero. As the court noted, privilege requires a reasonable expectation of privacy; a ToS that permits data mining destroys that expectation.

3. The “Alchemy” of Privilege

A dangerous misconception among clients (and some lawyers) is that sending a document to a lawyer retroactively shields it. Heppner forcefully rejects this.

The defendant argued that because he shared the AI outputs with his counsel to discuss strategy, they became privileged. Judge Rakoff called this a legal impossibility:

“[I]t is black-letter law that non-privileged communications are not somehow alchemically changed into privileged ones upon being shared with counsel” (p. 8).

The Takeaway: Practitioners must educate clients immediately: do not “prep” for our meeting by chatting with a bot. If a client creates a timeline of events using a public AI tool and then emails it to counsel, the email is privileged, but the underlying chat session logged on the AI server is not. That session remains a discoverable record of the client’s raw, unfiltered thoughts and admissions—a potential goldmine for opposing counsel.

4. No “Functional Equivalent” Protection for AI

Litigators often extend privilege to non-attorneys (accountants, interpreters) who are necessary to facilitate the attorney-client representation, known as the Kovel doctrine. Heppner draws a bright line: AI does not qualify.

The court reasoned that recognized privileges require “a trusting human relationship,” specifically with a professional who “owes fiduciary duties and is subject to discipline” (p. 6). The opinion bluntly states:

“Heppner does not, and indeed could not, maintain that Claude is an attorney” (p. 5).

The Takeaway: Do not rely on Kovel or “agency” theories to protect AI interactions unless the tool is a completely closed loop under the attorney’s direct control. You cannot deputize a public algorithm as a member of the legal team. If you need an AI analysis to be privileged, it must be generated by the legal team using secure tools, not by the client using public ones.

5. Distinction from the Shih Analysis

Defense counsel attempted to rely on Shih v. Petal Card, Inc., a prior S.D.N.Y. decision that extended work product protection to communications prepared without counsel’s specific direction. Judge Rakoff “respectfully disagree[d]” with Shih (p. 11).

He reaffirmed a stricter standard: the doctrine protects the lawyer’s mental processes, not the client’s independent anxieties or preparations.

“Because the AI Documents were not prepared at the behest of counsel and did not disclose counsel’s strategy, they do not merit protection as work product” (p. 12).

The Takeaway: Reliance on permissive, magistrate-level decisions regarding new technology is risky. Heppner signals a return to basics: the Work Product Doctrine is a narrow exception. Practitioners should assume that courts will apply these “longstanding legal principles” (p. 12) strictly, rather than expanding them to accommodate new tech behaviors.

Conclusion: The New Standard of Care

The Heppner memorandum concludes with a warning: “AI’s novelty does not mean that its use is not subject to longstanding legal principles” (p. 12).

For practitioners, this opinion apparently establishes a new standard of care for competence in the AI age:

1. Direct the Work: Never let a client use AI for legal tasks “on their own volition” if you want to claim privilege.

2. Read the TOS: If the tool “trains” on data, it is a privilege sieve.

3. Control the Tool: Privilege survives best when the lawyer holds the keys to the software.

The days of treating AI as a “magic box” (if there ever was) are over. It is now just another third party—and one that remembers everything.

Disclaimer: This is provided for informational purposes only and does not constitute legal or financial advice. To the extent there are any opinions in this article, they are the author’s alone and do not represent the beliefs of his firm or clients. The strategies expressed are purely speculation based on publicly available information. The information expressed is subject to change at any time and should be checked for completeness, accuracy and current applicability. For advice, consult a suitably licensed attorney and/or patent professional.

The official agency notice states, “Sponsored support staff are not permitted to insert a practitioner’s signature on any correspondence submitted to the USPTO” (Announcement, ¶ 1). The technical interface modification eliminated the option to insert registration numbers in the signature section for users lacking specific practitioner credentials.

The USPTO explicitly notes that “all signatures must be personally inserted by the named signatory” and that “sponsored support staff cannot insert the signature of a sponsoring practitioner or any other individual” (¶¶ 2, 4).

The notification forecasts further restrictions, indicating that “future Patent Center changes will remove the option for sponsored paralegals and other support staff to insert a signature” across the entire platform (¶ 6).

The agency provided a specific alternative workflow for law firms utilizing administrative personnel. The notice specifies that “sponsored support staff may prepare and save the Web ADS to permit a practitioner to insert their own signature” (¶ 5).

Some practitioners will notice that the USPTO provides an “alternative to the Web ADS is to submit a fillable PTO AIA/14 that a practitioner has signed.” The USPTO has long permitted an “S-signature” that “is a signature that is inserted between forward slash marks by the signer and is not a handwritten signature” (Fed. Reg. 2024-06126). Signing the PDF form with an S-signature obviates this manual sign-in and e-signature process, but could be a workflow shake-up for some firms.

These mechanical changes require attorneys to log into the system and authenticate documents individually, preventing the bulk submission processes that many high-volume practices formerly utilized.

The implementation of this signature procedure is fully active today. The procedural change functions as a direct administrative barrier preventing the utilization of automated signing protocols by artificial intelligence agents, such as Claude, ChatGPT, Gemini, and Open Claw.

Two years ago, in April 2024, the USPTO’s “Guidance on Use of Artificial Intelligence-Based Tools in Practice…” (Fed. Reg. 2024-07629) indicated a signature of a “person” is required:

Generally, all patent correspondence filed in the USPTO must bear a person’s signature. By including this signature, the individual inserting the signature or submitting the paper is certifying that the person’s signature appearing on the document was actually inserted by that person. In other words, a person, including a practitioner, must insert their own signature on the paper. “The requirement does not permit one person (e.g., a secretary) to type in the signature of a second person (e.g., a practitioner) even if the second person directs the first person to do so”

Patent professionals and industry observers detect the USPTO preparing for a substantial influx of AI-drafted patent applications. By mandating manual (electronic) signature insertion by the registered practitioner for such filings, the agency introduces deliberate friction into the prosecution process.

The restriction appears strategically formulated to throttle the volume of submissions. It forces attorneys to review and sign documents personally, effectively neutralizing fully autonomous AI filing pipelines that might otherwise submit thousands of applications without human intervention.

The patent office perceives this manual bottleneck as a necessary measure to keep the anticipated deluge of machine-generated filings manageable. Slowing down the attorneys serves as a primary method to control the influx of new applications.

This administrative update correlates strongly with other recent USPTO initiatives targeting procedural authenticity and accountability. A related policy shift involves the final rule requiring patent applicants and patent owners domiciled outside the United States to secure representation by a registered U.S. patent practitioner.

The Federal Register notice for the foreign counsel rule states the requirement aims to “enable the USPTO to more effectively use available mechanisms to enforce compliance... and enhance the USPTO’s ability to respond to false certifications” (Fed. Reg. 2026-05564).

Both the signature limitation and the foreign counsel requirement operate as structural methods to curb fraudulent filings. By affirming and/or establishing strict requirements for who can file and who must sign, patent administrators are constructing a defense against a predicted wave of unverified submissions that could overwhelm the examination corps.

The agency is applying friction points at the very beginning of the application lifecycle to verify the human origin and professional oversight of the submitted materials.

Before concluding an assessment of these regulatory shifts, analyzing the benefits, challenges, and risks of the Patent Center signature limitations provides a comprehensive perspective for IP owners and practitioners.

Benefits

Implementing strict manual (electronic) signature requirements yields distinct procedural advantages for the patent ecosystem. It preserves the integrity of the patent system by forcing a licensed human practitioner to authenticate each submission individually. This verification step limits the capacity for bad actors to flood the office with automated, low-quality filings generated entirely by artificial intelligence software.

The mandate allegedly protects the public interest by enforcing strict adherence to the rules of professional conduct. It holds practitioners directly accountable for the contents of the Web ADS and related correspondence, reducing the likelihood of careless errors or intentionally deceptive data entering the official record.

Challenges

The administrative burden on patent practitioners and their firms increases significantly. Attorneys must now manually execute routine documents that paralegals previously prepared and submitted autonomously. The separation of preparation and execution tasks creates logistical hurdles for high-volume law firms.

Practitioners might face slower filing pipelines and increased overhead costs associated with the constant interruption required to log into Patent Center and apply electronic signatures to staged documents. This division of labor disrupts established workflows that firms spent years optimizing for efficiency.

Risks

Introducing procedural friction carries inherent vulnerabilities for the legal community. The USPTO risks penalizing legitimate law firms by imposing heavy administrative workflows formulated to stop a minority of bad actors. The patent office might continue to add manual verification steps to combat AI automation. In that scenario, the cost of patent prosecution will inevitably rise.

The additional time required for compliance might hinder the economic feasibility of protecting inventions for smaller entities and independent inventors who rely on cost-effective legal services. Over-regulation in the filing process might stifle the exact innovation the agency exists to protect.

Conclusion

The Patent Center signature update represents a deliberate regulatory response to emerging technological capabilities in the legal sector. The USPTO is actively establishing administrative boundaries to manage the integration of artificial intelligence in patent prosecution.

Again, most practitioners will note that the USPTO provides an alternative to the Web ADS is to submit a fillable PTO AIA/14 that a practitioner has signed, e.g., with an S-signature.

Still, the requirement for manual electronic signatures could be a workflow shake-up for some firms and, paired with rules mandating U.S. representation for foreign applicants, indicates a deeper strategy to maintain strict human oversight over the patent system.

Patent Center has been far from bug free and implementing new software tweaks is always scary. It seems to have been implemented well enough thus far, and paralegals and staff should never have been signing electronic documents for the attorneys.

Patent professionals may have to adapt their internal workflows to accommodate these verification standards as the regulatory environment adapts to automated technologies.

Disclaimer: This is provided for informational purposes only and does not constitute legal or financial advice. To the extent there are any opinions in this article, they are the author’s alone and do not represent the beliefs of his firm or clients. The strategies expressed are purely speculation based on publicly available information. The information expressed is subject to change at any time and should be checked for completeness, accuracy and current applicability. For advice, consult a suitably licensed attorney and/or patent professional.

This initiative requires participating members to grant non-exclusive licenses to their AI foundation model patents, aiming to accelerate downstream development by mitigating the threat of costly legal friction.

SAIL is an active cross-license, where members cannot enforce against other members for infringing their AI foundation model patents. It acts as a non-aggression pact among active operating companies, but does include some protections in case an asset is transferred to, e.g., a patent assertion entity (PAE).

As of late April, members include Anthropic, Block, eBay, Figma, Genentech, IBM, Meta, Microsoft, and TD Bank Group.

To help IP strategists and counsel navigate the nuanced scope, asymmetrical withdrawal penalties, and hidden strategic risks of this agreement, here is an FAQ detailing what joining SAIL means for most companies and their patent portfolios.

Disclaimer: this is an academic discussion, based on hypothetical scenarios, and not intended to be legal or business advice.

Sail Membership Agreement

119KB ∙ PDF file

Download

Download

SAIL Foundation & Membership Basics

Q: What is the SAIL Foundation?

A: The SAIL Foundation is “the first organization dedicated to safeguarding AI innovation through a collaborative patent network” (p. 1). It “creates a collaborative licensing zone for the development of AI foundation models” to clear the way for rapid AI development (p. 1).

Q: Who are the members of SAIL?

A: The initiative is led by “founding board members Anthropic, Genentech (a member of the Roche Group), IBM, Meta, and Microsoft, and board observers eBay and TD Bank Group” (p. 1). Additionally, “Block and Figma have also joined as members” (p. 1).

Q: How much does it cost to join?

A: The dues are “$25,000/year (established company rate, discounts available for smaller companies” (p. 15). Failure to pay these fees “within sixty (60) days will result in termination of Membership” (p. 6).

Scope of the License

Q: What exact rights am I granting and receiving by signing this agreement?

A: Each member grants every other member a “worldwide, royalty-free, non-exclusive, non-sublicensable, and non-transferable license… under the Licensed Patents to make, have made, use, import, sell, offer for sale, lease, or otherwise distribute or provide Covered AI Technologies” (p. 1).

Q: Does this mean I am getting a license to every patent owned by companies like Microsoft and Meta?

A: No. You are only getting a license to their “Licensed Patents,” which the agreement strictly defines as the subset of patents that “would, in the absence of a license, be infringed by making, having made, using, importing, selling, offering for sale, leasing, or otherwise distributing or providing Covered AI Technology” (p. 10).

Q: What exactly is defined as a “Covered AI Technology”?

A: Covered AI Technologies include “(a) any Foundation Model; and (b) any software and services specifically designed or substantially modified to primarily serve the purpose of (i) training, fine-tuning, or adaptation of Foundation Models; (ii) testing, verification, validation, or monitoring of Foundation Models… (iii) integrating and/or interoperating with a Foundation Model… or (iv) implementing or enhancing safety mechanisms, capability controls, or oversight” (p. 8).

Q: I build end-user apps powered by AI. Is my application covered by this license?

A: No. The agreement explicitly states that “Covered AI Technologies exclude (x) software products and services built on Foundation Models, including but not limited to end user applications… and (y) hardware infrastructure” (pp. 8, 9).

Q: Does this agreement give me the right to use other members’ proprietary code or trade secrets?

A: No. The agreement contains a strict reservation of rights: “Except as expressly set forth in Sections 1 and 2 no license or right under any Patents or other intellectual property is granted by this Agreement, whether by implication, estoppel, or otherwise” (p. 1).

Press Release Holder

55.8KB ∙ PDF file

Download

Download

Risks, Terminations, & Suspensions

Q: What happens if I have already infringed on a member’s AI patent in the past?

A: The agreement grants a backward-looking “release of any and all claims, liabilities, and damages… for all infringement of Licensed Patents to the extent such claims are based on a Licensee’s Covered AI Technology and acts prior to the Effective Date” (p. 1).

Q: Under what circumstances can my license be suspended by another member?

A: A licensor can invoke a defensive suspension if you file “one or more patent infringement claims… against the Licensor or its Affiliates that are based on Licensor’s distribution or use of Covered AI Technologies” (p. 3). You can also be suspended if you initiate “a proceeding challenging the validity, patentability or enforceability of a Licensed Patent” (p. 3).

Q: If I join the SAIL Foundation and later decide to leave, do I get to keep my licenses?

A: Generally, no. The inbound “Licenses granted to a Foundation Member or its Affiliate… shall terminate effective as of the Exit Date” (p. 4). However, there is an asymmetrical penalty for leaving: your outbound patents “shall remain Licensed Patents and will remain and continue to be licensed… following the Exit Date to all Licensees existing as of the Exit Date” (p. 4). The only exception is if you have paid annual fees for at least three years, in which case you retain rights to patents with a priority date prior to your withdrawal (p. 4).

Comparisons to Other Patent Organizations

Q: Is SAIL like the LOT Network, OIN, RPX, Via Licensing, or Unified Patents?

A: While they all seek to reduce patent friction, their mechanisms are distinct.

• Unlike LOT Network: LOT is a defensive pact that only triggers if a patent is sold to a troll. SAIL is an active cross-license where members immediately grant “present, fully vested and irrevocable” licenses to each other for AI foundation models (p. 1).

• Unlike RPX & Unified Patents: RPX buys patents off the open market, and Unified Patents litigates to invalidate patents owned by companies asserting those assets (alleged “trolls”). SAIL does neither; it is purely a “collaborative patent network” and “shared patent commons” where operating companies pool their own IP (p. 1).

• Unlike Via Licensing: Via creates royalty-bearing patent pools to make money. SAIL is explicitly a “royalty-free” licensing pool designed to create a safe zone for development, not to generate licensing revenues (p. 1).

• Like the Open Invention Network: SAIL is essentially adopting OIN’s playbook, but applying it to AI instead of open-source software. Both are technology-specific cross-licenses. Members grant royalty-free, non-exclusive licenses to one another. Just as OIN defines a specific “Linux System” boundary for its license, SAIL draws a strict boundary around “Covered AI Technologies”. While OIN is generally free to join and does not permanently encumber patents if you leave, SAIL requires annual dues (e.g., $25,000/year) and features an asymmetrical withdrawal clause—if you leave SAIL, you lose your inbound protection but your previously granted outbound licenses remain in effect permanently (similar to LOT, but only applicable to specific subject matter).

Organizational Structure & Operations

Q: How is Jamster Capital involved?

A: Jamster Capital LLC is a boutique advisory firm that provides consulting and transaction services focused on intellectual property, founded by John Amster. Amster previously co-founded and led RPX Corporation to an $850 million IPO and served as CEO of Rowan Patents. Jamster Capital is not a member granting licenses, but rather the administrative operator. They have been “engaged by the Board of Directors to run the day-to-day operations of the Foundation, including member outreach, governance, marketing/communications, licensing guidance and accounting” (p. 2).

Q: What if the SAIL Foundation dissolves or goes bankrupt?

A: The licenses granted between members are “present, fully vested and irrevocable” and are intended to “run with the Licensed Patents to which they pertain and be binding on subsequent owners and licensees” (p. 1). Furthermore, if a licensor goes bankrupt, the agreement is treated as an executory contract under Section 365(n) of the U.S. Bankruptcy Code, allowing the licensee “to retain its license rights under this Agreement” (p. 6).

Q: Is this massive patent pool an antitrust issue?

A: The agreement includes standard safeguards used by pro-competitive patent pools to avoid antitrust scrutiny. Specifically, the licenses granted are “non-exclusive” (p. 1). Furthermore, members agree that this agreement “does not reflect a royalty that any Foundation Member or its Affiliate might otherwise have negotiated,” preventing it from being used as a price-fixing tool in outside disputes (p. 5).

Strategic Value for Potential Licensees

Q: Do I have to own a certain number of patents to join?

A: No. There is no minimum patent requirement mentioned in the agreement. You are required to grant a license to “each of its Subject Patents,” meaning whatever patents you do own, even if that number is currently zero (p. 1).

Q: How should I know if joining this will help my company?

A: This likely hinges on your company’s core business model and patent portfolio.

• It helps you if: You are building, training, or fine-tuning foundation models and need “freedom of action… to confidently invest” without fear of being sued by the massive portfolios of tech giants like Meta, IBM, or Microsoft (p. 1).

• Example: Imagine you are a well-funded startup training a new multimodal foundation model from scratch. Doing so might inadvertently infringe on fundamental model-training patents held by Microsoft or IBM. By paying the annual fee and joining SAIL, you instantly receive a royalty-free license to those patents for the purpose of training your model (p. 1). It acts as a relatively cheap, incredibly powerful insurance policy.

• It may hurt you if: You are an AI research firm that plans to monetize your own AI patents by asserting them against competitors or charging royalties, because the agreement requires you to grant a “royalty-free” license to all other members (p. 1). Furthermore, if you only build end-user applications on top of models, your core products are excluded from the license’s protection (pp. 8, 9).

• Example: Imagine your company builds an AI-powered CRM application (an end-user app) and you happen to own a highly valuable patent on a specific method for fine-tuning LLMs. If you join SAIL, you are forced to give Meta, IBM, and Anthropic a free license to use your fine-tuning patent (p. 1). However, because your actual product is just an end-user application built on top of a model, your CRM product is completely excluded from SAIL’s protective umbrella (pp. 8, 9). You might give away your crown-jewel IP for free and get essentially zero defensive value in return.

Disputes & Enforcement

Q: Are there any provisions about disputes between members or covered patents and where they must take place? Who determines if the patent is covered?

A: Yes, the agreement includes a strict forum selection and governing law clause. Any action or proceeding arising out of or relating to the agreement “shall be brought exclusively in the state or federal courts located in the State of Delaware,” and members irrevocably submit to that jurisdiction (p. 7). Furthermore, the agreement is interpreted and enforced strictly under Delaware law, without reference to choice of law principles (p. 7).

Regarding who determines if a patent is covered, the agreement does not establish an internal arbitration panel or technical tribunal. It also does not grant the Foundation Administrator the authority to adjudicate patent scope (p. 9). The Foundation Administrator’s role is strictly administrative, limited to publishing notices, handling withdrawals, and managing the Foundation Website (p. 9).

Because there is no mandatory arbitration for technical disputes, if two members disagree on whether a specific patent reads onto a “Covered AI Technology”—and thus qualifies as a “Licensed Patent” (p. 10)—that highly technical determination will ultimately have to be litigated and decided by a judge or jury in a Delaware court (p. 7).

Timing & Strategic Entry

Q: Since the agreement absolves past infringement, shouldn’t I just wait until a member threatens comes a-knockin’ before paying to join?

A: While it is true that the agreement grants a “release of any and all claims, liabilities, and damages… for all infringement of Licensed Patents… prior to the Effective Date” (p. 1), waiting on the sidelines is a highly dangerous strategy for two key reasons:

1. The Patent Transfer Vulnerability: The press release explicitly warns that “Inevitably, some assets will fall into the hands of assertion entities who will enforce them” (p. 2). If a SAIL member sells a patent to a non-member (like a patent troll) before you join, you cannot get a license to that patent through SAIL. However, if you join before the patent is sold, the agreement dictates that all licenses “run with the Licensed Patents” (p. 1) and any transfer “shall be subject to the Licenses” (pp. 1-2). Joining early locks in your protection; waiting leaves you exposed to third-party buyers.

2. Freedom to Operate and Due Diligence: Operating under the silent threat of litigation means you lack the “freedom of action… to confidently invest in technologies” (p. 1). If you are seeking venture funding or looking to be acquired, sophisticated investors will identify your unprotected use of foundation models during IP due diligence. That uncertainty can severely impact your valuation or kill a deal entirely long before a SAIL member ever officially “comes knocking.”

• Example: Imagine Meta decides to prune its massive AI portfolio and sells 50 older foundation model patents to an aggressive patent troll. If you had already paid your $25,000 and joined SAIL, your license to those 50 patents is fully vested and travels with the patents (p. 1). When the troll comes knocking, you simply show them your license and they go away. However, if you were waiting on the sidelines, you are out of luck. Because the troll is not a SAIL member, you cannot join SAIL to get a retroactive release against them. You tried to save $25,000 (per year) and now you are facing millions of dollars in federal litigation costs and damages.

Considerations for Patent Buyers & Assertion Entities

Q: What should I consider if I’m a PAE interested in acquiring AI-related patents from a member company?

A: As a PAE, your entire business model relies on the ability to enforce patents against operating companies. The press release acknowledges that “inevitably, some assets will fall into the hands of assertion entities who will enforce them” (Press Release, p. 2). However, buying patents from a SAIL member requires rigorous due diligence, as those assets may come heavily encumbered.

Here are the critical factors you must consider before acquiring an AI patent from a SAIL member:

1. The “Run with the Patent” Encumbrance: You are not buying a clean asset. The agreement dictates that all licenses granted “are intended to and shall run with the Licensed Patents to which they pertain and be binding on subsequent owners and licensees” (License, p. 1). This means the patent remains permanently licensed to every company that was a SAIL member at the time of the transfer.

2. Mandatory Acceptance of Terms: You cannot simply buy the patent and ignore the SAIL encumbrance. The agreement legally binds the seller to ensure that “any transferee shall agree that the transfer is subject to Licenses granted in this Agreement” (License, p. 2).

3. The “Covered AI Technology” Subject Matter Check: Before you finalize the acquisition, you must carefully map the claims of the patent to the agreement’s definitions. A patent only becomes an encumbered “Licensed Patent” if it would be infringed by making, using, or distributing “Covered AI Technology” (License, p. 10). If the patent specifically claims the architecture or training of a “Foundation Model” (License, p. 8), it is encumbered. However, if the patent’s claims are limited to “end user applications” (License, pp. 8-9) or “hardware infrastructure” (License, p. 9), those are explicitly excluded from the license scope, meaning you could still freely assert those specific claims against SAIL members.

4. Severely Reduced Target Landscape: Because the license runs with the covered patent, you will be completely barred from asserting that patent against any current SAIL Foundation member for their use of “Covered AI Technologies.” Your addressable market for litigation or licensing demands on those specific claims is instantly reduced to non-members.

• Example: Imagine you are a PAE evaluating a portfolio of 10 AI patents owned by Meta (a founding SAIL member). If you purchase those patents, you might plan to monetize them by suing IBM, Microsoft, and Anthropic. During due diligence, you must check the claims. If 8 of those patents cover core foundation model training techniques (License, p. 8), they are encumbered by a “present, fully vested and irrevocable” license to those SAIL members (License, p. 1). You cannot sue IBM or Microsoft for using those 8 patents.

• However, if the remaining 2 patents exclusively cover AI hardware configurations (License, p. 9), those are not “Covered AI Technologies.” You could acquire those 2 patents and successfully assert them against Microsoft and IBM, because SAIL does not grant them a license for hardware.

Q: If I buy a disputably covered patent from a member, am I stuck enforcing it in Delaware?

A: If you are trying to enforce it against a current SAIL Foundation member, almost certainly yes. While a patent buyer or Patent Assertion Entity (PAE) might strongly prefer to file an infringement lawsuit in a traditionally patent-friendly jurisdiction (like the Eastern or Western Districts of Texas), the SAIL agreement heavily complicates your choice of venue.

When you acquire a patent from a member, the agreement legally binds the seller to ensure that “any transferee shall agree that the transfer is subject to Licenses granted in this Agreement” (p. 2). If you sue a SAIL member in Texas and they raise the defense that the patent’s subject matter is covered by the SAIL license, that specific dispute inherently requires interpreting the agreement.

The agreement explicitly mandates that “[a]ny action or proceeding arising out of or relating to this Agreement shall be brought exclusively in the state or federal courts located in the State of Delaware” (p. 7). Therefore, the defending member will almost certainly leverage this clause to successfully transfer the case to Delaware, or they will preemptively file a Declaratory Judgment action against you in Delaware to have a judge rule on whether the patent is covered.

Strategic Exception: If you are enforcing that exact same acquired patent against a non-member, the SAIL agreement’s defensive licenses—and thereby its Delaware forum selection clause—do not apply to that specific dispute. In that scenario, you are free to file the lawsuit wherever standard federal venue laws allow.

Final Thoughts

Ultimately, joining the SAIL Foundation is a calculated trade-off between sacrificing potential licensing revenue and securing an essential defensive shield within the rapidly evolving AI landscape.

While the asymmetrical exit penalties and loss of assertion rights against fellow members pose significant long-term constraints, these risks must be weighed against the immediate protection from predatory patent transfers and costly litigation.

For companies prioritizing rapid innovation over patent monetization, SAIL offers a powerful insurance policy, provided you’re willing to play in a shared sandbox where the fences are high and the exits are steep.

Disclaimer: This is an academic discussion, based on hypothetical scenarios, and not intended to be legal or business advice. This is provided for informational purposes only and does not constitute legal or financial advice. To the extent there are any opinions in this article, they are the author’s alone and do not represent the beliefs of his firm or clients. The strategies expressed are purely speculation based on publicly available information. The information expressed is subject to change at any time and should be checked for completeness, accuracy and current applicability. For advice, consult a suitably licensed attorney and/or patent professional.

When legal privilege does not shield these records, the communications become direct evidence or provide opposing counsel with a detailed guide to finding adverse facts. In intellectual property law, an innocuous query or internal memorandum can quickly become the central focus of a judge or jury.

The recent federal court memorandum in United States v. Heppner demonstrates the severe consequences of unprivileged digital interactions. The government seized 31 chat logs generated between the defendant and the artificial intelligence platform “Claude.”

The defendant attempted to shield these logs by claiming attorney-client privilege and the work-product doctrine. Judge Jed S. Rakoff of the Southern District of New York rejected both arguments.

The court noted that the user agreement permitted the platform to collect data and “disclose personal data to third parties” (p. 6). The court found the user “could have had no ‘reasonable expectation of confidentiality in his communications’” (p. 7). By communicating with a public platform, the defendant generated a discoverable, unprivileged record of his actions.

The Heppner case involved a criminal defendant attempting to shield evidence of white-collar offenses; nevertheless, the underlying legal principles apply directly to intellectual property practice. The court established that a user acting independently cannot claim they were acting as an agent of a lawyer. The judge wrote, “[B]ecause the AI Documents were not prepared at the behest of counsel and did not disclose counsel’s strategy, they do not merit protection as work product” (p. 12).

A common practice in research departments involves inventors using digital tools to draft technical summaries, forwarding those summaries to outside counsel for review.

The court in Heppner explicitly rejected the theory that forwarding a document to a lawyer retroactively shields the original interaction. The judge clarified that “non-privileged communications are not somehow alchemically changed into privileged ones upon being shared with counsel” (p. 8). The initial generation of the text remains discoverable.

The Roadmap to the Smoking Gun

In commercial litigation, unprivileged communications serve as a roadmap to the smoking gun, if not the evidence itself. Electronic discovery sweeps capture vast amounts of internal data. Search algorithms flag conversations involving specific patent numbers, competitor names, or technical keywords.

If a chat log or standard email lacks privilege, opposing counsel gains immediate access to, e.g., thought processes of the inventors and executives. This opens a door.

The lack of privilege means opposing counsel does not need to guess where a company’s vulnerabilities lie. A single unprivileged document might indicate that an engineering team struggled to differentiate their new product from an existing patent.

Armed with this document, litigators know exactly which engineers to depose and exactly which questions will yield damaging answers. The document provides the exact coordinates for a smoking gun.

Pre-Filing and Prosecution Hazards

For inventors and in-house counsel, the period before filing a patent application carries significant evidentiary exposure. Internal discussions and prompts entered into generative models often contain candid assessments of a new technology. This is a very delicate time to be risking inputting details into any AI platform or search without consulting proper counsel.

For instance, a primary concern involves admissions regarding prior art and the scope of the invention. An inventor detailing the limitations of their work compared to existing solutions might inadvertently create a record that narrows the perceived scope of the invention.

Such documentation provides adversarial parties with ready-made arguments against patentability or broad claim construction.

Dates of conception and reduction to practice remain highly vulnerable to unprivileged disclosures. Informal records discussing the timeline of an invention can contradict later assertions made under oath during interference proceedings or litigation.

If an engineer types a prompt asking a chatbot to help solve a technical hurdle on a specific date, that log may establish that the invention was not fully reduced to practice at that time—or that the “inventor” was relying on the AI for RTP.

Patent practitioners must always evaluate the duty of candor owed to the United States Patent and Trademark Office. If an unprivileged chat log or internal email reveals that an inventor discovered a highly relevant piece of prior art, and that reference fails to appear in an Information Disclosure Statement, the patent owner faces severe allegations of inequitable conduct.

A finding of inequitable conduct renders the entire patent unenforceable. The digital record can serve as direct proof of knowledge of the party and, potentially, an intent to deceive the patent office.

Litigation, Claim Construction, and Willfulness

The evidentiary risks extend well into the enforcement phase of a patent’s lifecycle. During litigation, the characterization of an invention in unprivileged documents heavily influences claim construction.

A court frequently looks at how the inventors described their own technology internally to determine the ordinary meaning of disputed claim terms. If an inventor used narrow, specific language to describe a feature in a chat prompt or internal email, opposing counsel will use that document to argue for a narrow claim interpretation, potentially allowing competitors to design around the patent.

Unprivileged records create massive exposure regarding willful infringement. Formulating a prompt that asks a chatbot to compare an internal product design against a specific competitor patent establishes direct knowledge of that patent.

In a subsequent infringement lawsuit, those logs may demonstrate that the defendant knew about the IP asset(s) and analyzed potential vulnerabilities.

Asking the wrong artificial intelligence platform whether an invention infringes on a specific patent may generate a discoverable admission of concern.

The court in Heppner emphasized that privilege requires “a trusting human relationship” with a licensed professional subject to ethical discipline (p. 6). An algorithm is not a licensed attorney; accordingly, the interaction holds zero legal privilege status. This documentation directly supports claims for enhanced damages based on willful infringement, exposing the company to treble damages.

Logs detailing research into potential infringement give away key weaknesses. Creating evidence-of-use charts without direct supervision from legal counsel generates unprivileged material. Opposing counsel reviewing these charts gains immediate insight into the internal legal theories and perceived weaknesses of the patent owner’s case.

Analysis

Analyzing the integration of generative tools and informal digital communication in intellectual property development requires balancing productivity against legal exposure.

Modern communication platforms and generative artificial intelligence accelerate the pace of innovation. Inventors can draft technical specifications, summarize voluminous prior art, and refine complex engineering problems at unprecedented speeds.

These tools offer significant cost savings and allow research teams to iterate designs rapidly. The optimism surrounding these capabilities is well-founded; they provide substantial technical assistance for sophisticated engineering tasks.

They lower the barrier to entry for drafting initial invention disclosures, allowing engineering teams to document their ideas efficiently before engaging costly legal counsel.

The primary difficulty lies in training personnel to recognize the boundaries of legal privilege. Engineers and researchers often view interactions with digital tools as private brainstorming sessions, similar to writing in a locked laboratory notebook. Correcting this assumption requires comprehensive policy implementations and constant training.

Managing data retention presents another significant hurdle. Organizations must track where technical data is entered, how long it is stored, and who has access to it, complicating information technology governance.

Establishing a workflow where inventors only conduct prior art searches or infringement analyses under the specific, written direction of legal counsel slows down the research process and introduces administrative friction.

The most severe threat is the total waiver of confidentiality. Transmitting data to third-party platforms destroys reasonable expectations of privacy. For intellectual property owners, this destruction immediately threatens trade secret protection.

Maintaining a trade secret requires an organization to take reasonable measures to keep the information secret. Entering proprietary source code, chemical formulas, or customer data into a standard, public-facing generative platform actively defeats those reasonable measures.

Once the court determines that the user surrendered the expectation of privacy, the information legally ceases to function as a trade secret.

The discoverability of internal doubts, alternative design failures, and prior art knowledge provides adversarial litigators with a highly targeted index of a company’s vulnerabilities. The lack of a functional attorney-client relationship with automated platforms guarantees that any legal question asked by an inventor becomes public record during discovery.

Conclusion

The intersection of advanced digital tools and intellectual property law demands strict adherence to established legal principles. The Heppner decision acts as a stark warning that unprivileged communications reliably guide opposing parties directly to adverse facts.

For inventors and IP owners, the new protocol is simple but strict:

1. Assume Visibility: Treat every prompt box as a public bulletin board.

2. Verify the Tool: Only input proprietary data into AI tools that have been vetted by legal counsel and have “zero-retention” policies confirmed by contract.

3. Wait for Direction: If the task involves legal analysis (like prior art charting or infringement checks), wait for specific instruction from counsel on which tools to use and how to use them.

Patent practitioners and intellectual property owners should operate under the assumption that informal digital records will face scrutiny in future legal proceedings.

Protecting innovation requires implementing strict data entry protocols and requiring explicit, documented direction from legal counsel before analyzing patents or prior art.

Prudent management of digital footprints remains the most effective strategy for preventing casual interactions from becoming damaging evidence.

Disclaimer: This is provided for informational purposes only and does not constitute legal or financial advice. To the extent there are any opinions in this article, they are the author’s alone and do not represent the beliefs of his firm or clients. The strategies expressed are purely speculation based on publicly available information. The information expressed is subject to change at any time and should be checked for completeness, accuracy and current applicability. For advice, consult a suitably licensed attorney and/or patent professional.

This “Guidance” document functions as a set of guiding principles based on existing professional responsibility obligations. The California Supreme Court recently directed the State Bar to incorporate these principles into the formal rules and address autonomous agentic systems. These proposed rules are a bit more strict.

The resulting proposed amendments to the Rules of Professional Conduct compel COPRAC to reevaluate and revise the 2023 Practical Guidance ahead of the May 2026 Board of Trustees meeting. Here is a good guess of what will change.

Scope Expansion: Integrating Agentic Technology

The 2023 AI Guidance focuses almost exclusively on generative applications. The anticipated revisions will expand this scope to cover agentic systems. Agentic models operate autonomously, executing multi-step workflows without continuous human prompting. This access and freedom presents greater risk for those in the legal field.

COPRAC will likely update the introductory sections of the guidance to define agentic systems and outline their specific operational hazards compared to reactive generative tools.

2023 Generative Ai Practical Guidance

135KB ∙ PDF file

Download

Download

Conflict 1: The New Disclosure Standard

• 2023 Practical Guidance: “The lawyer should consider disclosure to their client that they intend to use generative AI in the representation...” (Guidance, p. 4).

• Proposed Rule 1.4 Amendment: The new Comment [5] states that a lawyer “must communicate sufficient information regarding the use of technology to permit the client to make informed decisions” (Proposed Amendments).

• Analysis: The 2023 Guidance frames client communication as an optional consideration. The proposed amendment converts this into a strict mandate. Attorneys can no longer rely on internal discretion; they face an affirmative duty to disclose technology usage that presents material risks or affects the scope of representation.

Conflict 2: The Supervisory Suggestion

• 2023 Practical Guidance: “Managerial and supervisory lawyers should establish clear policies regarding the permissible uses of generative AI and make reasonable efforts to ensure that the firm adopts measures that give reasonable assurance that the firm’s lawyers and non lawyers’ conduct complies with their professional obligations when using generative AI...” (Guidance, p. 3).

• Proposed Rule 5.1 Amendment: The proposed modification dictates that “managerial lawyers must make reasonable efforts to establish internal policies and procedures” (Proposed Amendments).

• Analysis: The 2023 text presents policy creation as a best practice recommendation. The proposed rule elevates this to a formal ethical obligation, subjecting firm leadership to disciplinary action for failing to implement structured, written policies governing automated systems.

Conflict 3: The Third-Party Data Standard

• 2023 Practical Guidance: “A lawyer who intends to use confidential information in a generative Al product should ensure that the provider does not share inputted information with third parties...” (Guidance, p. 2).

• Proposed Rule 1.6 Amendment: The proposal defines “reveal” as “exposing confidential information to technological systems... where such exposure creates a material risk” (Proposed Amendments).

• Analysis: The 2023 Guidance focuses heavily on reviewing vendor terms of service. The proposed amendment establishes a stricter baseline where the mere act of exposure constitutes a breach if a material risk exists. Practitioners face a higher burden of vetting system architecture before inputting sensitive client data.

Conflict 4: The “Starting Point” Assumption

• 2023 Practical Guidance: “Al-generated outputs can be used as a starting point but must be carefully scrutinized.” (Guidance, pp. 2-3).

• Proposed Rule 1.1 Amendment: The Supreme Court directive specifically highlights “agentic artificial intelligence tools, which can enable systems to autonomously perform tasks or workflows without human prompting” (Proposed Amendments, Background). Comment [2] states a lawyer “must independently review, verify, and exercise professional judgment regarding any output” (Proposed Amendments).

• Analysis: Characterizing machine outputs strictly as a “starting point” fails to accurately capture the function of agentic systems that complete end-to-end workflows autonomously. COPRAC will likely revise this language to emphasize the mandatory independent verification of finalized, automated tasks before execution.

Conflict 5: The “Accuracy” Versus “Existence” Verification

• 2023 Practical Guidance: “A lawyer must review all generative AI outputs, including, but not limited to, analysis and citations to authority for accuracy before submission to the court...” (Guidance, p. 4).

• Proposed Rule 3.3 Amendment: The new Comment [3] dictates an “obligation to verify the accuracy and existence of cited authorities, including ensuring no cited authority is fabricated...” (Proposed Amendments).

• Analysis: The 2023 guidance requires a general review for accuracy. That’s not likely strong enough anymore. The proposed amendment directly addresses the specific hazard of algorithmic hallucinations by adding an affirmative duty to verify the physical existence of a case. Reviewing a document for typographical or formatting correctness falls short. Practitioners bear a specific ethical obligation to confirm that an autonomous system did not fabricate the cited authority entirely.

Conclusion

The transition from general guidance to binding ethical rules marks a maturing regulatory framework for legal technology. Still, it’s hard to believe that California’s AI Guidance has sat untouched since 2023. So much has happened.

Developments in California—especially regarding technology implementation—carry significant weight for legal professionals nationwide. The state’s proposed amendments establish a strict baseline for the independent verification of agentic systems and continuous informed client consent.

Ethics committees in New York, Illinois, New Jersey, Pennsylvania, and potentially federal agencies like the United States Patent and Trademark Office will likely follow California’s lead. These jurisdictions possess a high probability of adopting similarly rigorous standards for autonomous workflows. Legal practitioners across the country should evaluate these proposals to prepare for future compliance obligations.

IP attorneys, litigators, and in-house counsel face a permanent shift in daily task execution and must adapt internal workflows to satisfy these strict verification and disclosure standards. COPRAC’s upcoming revisions solidify the expectation that human professional judgment dictates all automated processes.

The public comment period remains active until May 4, 2026. Practitioners possess an opportunity to shape the final regulations by submitting feedback through the online Public Comment Form.

COPRAC will present further modifications to the existing Practical Guidance at the May 2026 Board of Trustees meeting.

Disclaimer: This is provided for informational purposes only and does not constitute legal or financial advice. To the extent there are any opinions in this article, they are the author’s alone and do not represent the beliefs of his firm or clients. The strategies expressed are purely speculation based on publicly available information. The information expressed is subject to change at any time and should be checked for completeness, accuracy and current applicability. For advice, consult a suitably licensed attorney and/or patent professional.

The agency expects the process to “permit the USPTO to efficiently and effectively address the recent increased volume of ex parte reexamination requests by providing examiners with additional information useful for making the SNQ determination” (Announcement, ¶ 2).

The change provides patent examiners with additional context before they determine if a third-party request raises a substantial new question of patentability.

Why Ex Parte Reexamination?

The February 2026 statistics reveal a massive surge in ex parte reexamination filings.

In FY 2025 the Office saw 491 requests, up from 417 in 2024 and 288 in 2023, marking a sharp resurgence and the highest annual volume for these proceedings since 2012.

Even more shocking, in Q1 of FY 2026 (October through December 2025), the USPTO reported 223 new filings, a historic quarterly pace that exceeds the total annual requests received in any single year between 2016 and 2020.

Based on that, it appears that the USPTO may be facing a significant resource logjam as inventory levels swell and average pendency times for these proceedings climb past 21 months.

Consequently, patent practitioners are shifting their approach to view these proceedings as a key strategic pillar rather than a mere fallback, even with the USPTO’s tightening rules on anonymity and estoppel.

New Procedures

Under 35 U.S.C. 303(a), the Director must determine whether a request raises a substantial new question of patentability “within three months following the filing of a request for reexamination” (OG, p. 1). During the initial three-month window, the Office historically lacked a mechanism to receive input from the patent owner. The Office noted that under prior practice, it “has not been able to obtain the benefit of relevant information from patent owner prior to making the substantial new question determination” (OG, p. 1).

Patent owners had to remain silent until after an order granting reexamination to file a statement. The statutory blackout period meant that examiners formulated their initial opinions relying exclusively on the third-party requester’s framing of the prior art and prosecution history. The absence of a formal rebuttal mechanism often allowed requests with questionable assertions to clear the preliminary threshold, forcing patent holders into protracted administrative proceedings.

Official Gazette Preorder SNQ Apr2026

1.19MB ∙ PDF file

Download

Download

The new USPTO framework alters the response timeline. Patent owners may submit a paper titled “Patent owner pre-order paper providing information useful in making the SNQ determination” (OG, pp. 1-2). The submission allows the patent owner to argue “why an argued teaching(s) in a proposed SNQ presented in a request for reexamination would not be considered important to a reasonable examiner” (Announcement, ¶ 1). The procedural tool introduces a preliminary adversarial element to a process that historically operated as a one-sided evaluation during its earliest phase.

The standard for finding a substantial new question remains unchanged. As referenced in the notice, the Manual of Patent Examining Procedure specifies that printed publications raise a substantial new question when “(1) the teaching of the patents and printed publications is such that a reasonable examiner would consider the teaching to be important in deciding whether or not the claim is patentable; and (2) the same question of patentability as to the claim has not been decided by the Office in an earlier concluded examination” (OG, p. 1).

According to the Official Gazette, the Federal Circuit previously observed that “the reexamination statute’s purpose is to correct errors made by the government and if need be to remove patents that never should have been granted” (OG, p. 1).

The Office placed strict parameters on the new pre-order submissions. The paper “must be limited to thirty (30) pages or fewer” and must align with the formal requirements of 37 CFR 1.52 (OG, p. 2).

Patent owners must file the response “as soon as possible, but no later than the date that is 30 days from the date of service of the request on the patent owner” (OG, p. 3). The 30-day period cannot be extended under any circumstances.

Substantive Matters

The content of the paper is restricted to arguments and facts demonstrating that the Office should maintain the original decision of patentability. The submission “should not address matters not raised in the request” (OG, p. 2). Practitioners must avoid statutory discretion arguments entirely.

The rules state the paper “specifically should not address why the USPTO should exercise discretion under 35 U.S.C. 325(d), which is taken up after determination of whether there is a substantial new question” (OG, p. 2).

Arguments that an alleged teaching is entirely new should be reserved for a later stage. The Office advised that such challenges “should not be included in a patent owner pre-order paper” considering the short timeframe for a determination (OG, p. 4).

This all points to jumping right into challenging the references within this tight 30-day window.

Patent owners can support their pre-order papers with evidence. Practitioners can file a declaration to support their arguments. A declaration “will not impact the page limit” (OG, p. 2). The Office restricts formatting tactics, stating it “will rely on arguments made in the paper and incorporation by reference will not be permitted” (OG, p. 2).

Other Procedures

Beyond content restrictions, the Office enforces strict service protocols. The patent owner pre-order paper must be served on the third-party requester in accordance with 37 CFR 1.248 (OG, p. 2). This rule limits service methods to electronic mail (provided the parties agree in writing), first-class mail, or direct hand delivery. Practitioners must verify the paper reflects this service. If a submission lacks proper documentation of service, the Office warned that the papers “may be refused consideration” (OG, p. 2).

Third-party requesters face heavy restrictions on replying to the patent owner’s pre-order paper. The Office reasons that the requester “has already filed its request, which has no page limitations” (OG, p. 2). In narrow circumstances, exceptions exist. Requesters may seek permission to file a response “to address alleged misrepresentations of fact or law or other improper arguments that materially impede the determination of a substantial new question” (OG, p. 2).

To utilize the exception, the requester must submit “a responsive paper accompanied by a grantable petition under 37 CFR 1.182 (and fee)” (OG, p. 2). If permitted, the reply is capped at ten pages and must be filed within 15 calendar days from the date of service of the patent owner’s paper (OG, p. 3).

The USPTO streamlined the administrative execution of the process. The Director used authority under 37 CFR 1.183 to automatically waive the provisions of 37 CFR 1.530(a) and the second sentence of 1.540 for timely filed pre-order papers (OG, p. 3).

A separate petition or fee for the waiver is not required. The Office framed the waiver as a response to the “increased volume of ex parte reexamination requests” and the need to gather information efficiently (OG, p. 3).

While the USPTO’s notice page indicates that the paper will be published on April 28th, it appears to go into effect on April 5th. Because this is framed as a waiver of procedural rules (specifically waiving 37 CFR 1.530(a) and part of 1.540 under the Director’s 1.183 authority), the USPTO has the flexibility to implement it for any new proceeding filed on or after that designated April 5 date, even if the formal publication happens weeks later.

Benefits

The implementation of the pre-order paper introduces distinct benefits for patent holders defending their intellectual property. The primary advantage is the opportunity to terminate a reexamination request before it officially begins. An early dismissal saves significant financial resources and limits periods of uncertainty regarding patent validity.

Permitting declarations provides an avenue to introduce expert testimony early in the proceedings. Factual assertions from an industry expert can quickly neutralize an alleged teaching that a reasonable examiner might otherwise find persuasive on a first read. Furthermore, the procedure is highly accessible from an administrative standpoint; the USPTO permits the filing “without filing a petition or paying a fee” and automatically waives the restrictive provisions of 37 CFR 1.530(a) (OG, pp. 1, 3).

The new mechanism empowers patent owners to take an active role rather than remaining bystanders for the first three months of the proceeding. A successful pre-order paper protects the commercial value of the underlying invention by removing the cloud of a pending patentability challenge.

Challenges

The procedure presents substantial procedural and logistical hurdles for patent practitioners. The 30-day deadline demands immediate action. Coordinating with inventors, securing expert witnesses, and analyzing complex prior art requires a highly efficient response upon receiving a request.

For most parties—especially small entities and independent inventors—retaining counsel and drafting a 30-page substantive response accompanied by a declaration within one month demands rapid mobilization of capital and personnel. The unextendable nature of the deadline adds intense pressure to the review process.

Additionally, the rules strictly limit the scope of arguments available. IP professionals must resist the urge to argue that a reference fails to constitute a “new” teaching, as the Office expressly warned that “such arguments should not be included” given the limited timeframe (OG, p. 4).

Similarly, practitioners are barred from raising 35 U.S.C. 325(d) discretionary arguments. Adhering to the strict formal requirements of 37 CFR 1.52 and ensuring proper service on the third-party requester under 37 CFR 1.248 leaves little margin for error in execution.

Risks

Several risks require careful evaluation before a patent owner submits a pre-order paper. Statements made in the submission become part of the permanent prosecution history. Prosecution history estoppel can limit claim scope in future or concurrent litigation. Committing to specific technical arguments or claim constructions early in the process creates rigid positions.

Procedurally, failing to comply strictly with the limitations poses a direct threat to the filing. Papers that do not meet the 30-page limit, formatting requirements, or service protocols “may be refused consideration” entirely (OG, pp. 2, 3).

Even worse, if a patent owner includes what a requester views as improper arguments or misrepresentations, the requester can petition to file a responsive paper (OG, p. 2). This scenario risks granting the challenger an additional opportunity to characterize the art before the examiner.

Finally, if the pre-order paper fails to prevent the institution of the reexamination, the patent owner has provided the third-party requester with a comprehensive preview of the defense strategy. The requester can analyze the expert declarations and arguments, allowing them to adapt their litigation tactics or prepare subsequent challenges with a clearer view of the patent owner’s theories. Practitioners must weigh the probability of successfully preventing institution against the long-term strategic implications of early disclosure. Submitting a pre-order paper is a highly strategic decision requiring a balancing of short-term relief and long-term enforceability.

Practitioner Takeaways

• Days 1–5 - Consider Service: Because the submission must be served pursuant to 37 CFR 1.248, practitioners and staff should immediately set a plan for service. For instance, counsel may want to seek written agreement from the requester for electronic mail service to avoid logistical delays and tracking uncertainties of relying on first-class mail or hand delivery.

• Days 1–30 - Rapid Resource Allocation: Upon receiving service of an ex parte reexamination request, patent practitioners face an unextendable 30-day window to file a pre-order paper. IP professionals should establish protocols for immediate mobilization of inventors and expert witnesses to meet this strict deadline.

• Days 1–30 - Strategic Use of Declarations: During the preparation period, counsel can leverage the procedure to submit expert declarations without impacting the 30-page limit—but patentees cannot incorporate arguments from the declaration by reference. Introducing factual assertions from industry experts early can potentially neutralize prior art arguments before a formal proceeding begins.

• 15 Days After Patentee Pre-Order Paper - Requester Reply: If a patent owner timely files a pre-order paper, the third-party requester has 15 calendar days from service to seek permission to file a responsive paper via a petition under 37 CFR 1.182. Patentees and counsel must ensure their pre-order submissions strictly comply with the rules to avoid granting the challenger an additional opportunity to characterize the art.

• Month 3: Assessing Defensive Previews: The Director will determine whether a substantial new question exists within three months following the request’s filing. If the pre-order paper fails to prevent institution at this phase, the patent owner has supplied the third-party requester with a preview of the defense strategy. IP owners and in-house counsel must assess whether the potential for early termination outweighs the risk of allowing challengers to adapt their litigation tactics.

• Long-Term: Prosecution History Estoppel: Regardless of the institution decision, statements made in the pre-order submission enter the permanent prosecution history. Attorneys must carefully weigh the probability of a successful early dismissal against the long-term impact of committing to specific technical arguments or claim constructions that could limit future enforceability.

Conclusion

The introduction of the pre-order paper marks a substantial shift in ex parte reexamination practice. The USPTO’s effort to integrate patent owner feedback earlier in the process reflects a desire for more informed decision-making during the substantial new question determination phase.

By evaluating arguments from both sides before instituting a formal proceeding, the Office aims to allocate examination resources more efficiently. This initiative flows with what Director Squires has been implementing: aiming to maintain the original decision of patentability unless there is a true mistake and reduce the backlog(s).

The biggest acceleration is that patent owners are effectively defending the validity of their claims over the cited art before the Office has even formally decided to reexamine them. The new rules effectively defer patentee arguments about the “discretion” and “newness” of the references.

Patent practitioners have to hustle to examine the art and work with an expert declarant—counsel must adjust their internal intake procedures to accommodate the strict 30-day filing window. The clock starts at service, but there will surely be questions about delay due to mail and/or Patent Center not running smoothly.

Patent owners face a new tactical decision when confronted with an ex parte reexamination request, requiring swift evaluation of the merits and risks of early engagement.

The process creates a formalized channel for early advocacy, reshaping the initial stages of post-grant patent challenges. Due to the tight timeframe, it won’t be without its initial hiccups.

Disclaimer: This is provided for informational purposes only and does not constitute legal or financial advice. To the extent there are any opinions in this article, they are the author’s alone and do not represent the beliefs of his firm or clients. The strategies expressed are purely speculation based on publicly available information. The information expressed is subject to change at any time and should be checked for completeness, accuracy and current applicability. For advice, consult a suitably licensed attorney and/or patent professional.

In United States v. Heppner, No. 25 Cr. 503 (JSR) (S.D.N.Y. Feb. 17, 2026), Judge Jed S. Rakoff of the Southern District of New York ruled that a criminal defendant’s interactions with the AI platform “Claude” were protected by neither attorney-client privilege nor the work product doctrine.

The decision, while not necessarily precedential, serves as a significant warning to the legal community regarding the confidentiality risks inherent in public generative AI tools.

Background of the Dispute

The controversy arose within a criminal securities fraud prosecution. The government charged Bradley Heppner, an executive at GWG Holdings, Inc., with multiple counts of fraud and falsifying corporate records (Memorandum, p. 2). Following Heppner’s indictment and arrest in November 2025, the FBI executed a search warrant at his residence, seizing various electronic devices.

Heppner Memorandum 02-17-2026

2.56MB ∙ PDF file

Download

Download

During the review of the seized materials, defense counsel asserted privilege over approximately thirty-one documents. These documents were not traditional emails to lawyers or handwritten notes; rather, they memorialized communications Heppner had with “Claude,” the generative AI platform operated by Anthropic (p. 3).

According to the defense, Heppner engaged with Claude after receiving a grand jury subpoena and understanding he was a target of the investigation. Heppner allegedly used the AI to prepare reports outlining his defense strategy and anticipating arguments regarding the facts and law (p. 3).

The defense argued these inputs incorporated information conveyed by counsel and were created for the express purpose of obtaining legal advice (pp. 3-4).

The government moved for a ruling that these “AI Documents” were not privileged. Judge Rakoff granted the motion from the bench during a pretrial conference and subsequently issued a written Memorandum explaining the court’s reasoning.

The Court’s Analysis: Unpacking the Privilege Claims

The court addressed two distinct legal theories: attorney-client privilege and the work product doctrine. The analysis for both turned on the nature of the relationship—or lack thereof—between a human user and an AI algorithm.

Attorney-Client Privilege: The Missing Elements

The court dismantled the claim of attorney-client privilege by applying the standard three-part test: the privilege requires a communication between a client and attorney, intended to be confidential, for the purpose of legal advice. Judge Rakoff found the AI Documents failed on all three fronts.

First, and perhaps most obviously, the court noted that “Claude is not an attorney” (p. 5). While this seems elementary, the court rejected the “functional equivalent” argument often used for non-attorney experts (such as accountants) who assist lawyers. The court observed that recognized privileges require “a trusting human relationship,” specifically with a professional bound by fiduciary duties and ethical rules. An AI platform satisfies none of these criteria.

Second, the court found that the communications were not confidential. This determination is particularly relevant for corporate counsel managing data security. Judge Rakoff pointed directly to Anthropic’s privacy policy, which users must accept. The policy stated that the company collects data on inputs and outputs to “train” the model and reserves the right to disclose data to third parties, including regulatory authorities (p. 6).

Because Heppner voluntarily disclosed his thoughts to a third-party platform that retains data in the ordinary course of business, he waived any expectation of confidentiality. As the court noted:

“Heppner could have had no ‘reasonable expectation of confidentiality in his communications’ with Claude. And the AI Documents are not like confidential notes that a client prepares with the intent of sharing them with an attorney because Heppner first shared the equivalent of his notes with a third-party, Claude” (p. 7).

Third, the court rejected the argument that Heppner used Claude to obtain legal advice. The opinion highlights that Claude itself creates a disclaimer stating, “I’m not a lawyer and can’t provide formal legal advice” (p. 8). Furthermore, the court emphasized that Heppner acted on his own volition. Had counsel directed Heppner to run specific searches on an AI tool to assist in legal representation, the analysis might have differed. However, because Heppner acted independently, his intent to eventually share the output with counsel did not retroactively privilege the initial interaction with the AI (p. 7).

The Work Product Doctrine: No “Zone of Privacy” for AI

The defense presented a more nuanced argument regarding the work product doctrine, which generally protects materials prepared in anticipation of litigation. Heppner argued that because he created these documents to aid his defense strategy after receiving a subpoena, they should be shielded.

Judge Rakoff disagreed, clarifying that the core purpose of the work product doctrine is to protect the attorney’s mental processes, not the client’s independent activities. The court relied on Second Circuit precedent establishing that the doctrine typically does not protect materials in a client’s possession unless they reflect the lawyer’s thinking or were prepared at the lawyer’s request.

The opinion explicitly states:

“The AI Documents do not merit protection under the work product doctrine because... they were nevertheless not ‘prepared by or at the behest of counsel,’... nor did they reflect defense counsel’s strategy” (p. 9).

The defense attempted to rely on Shih v. Petal Card, Inc., a 2021 decision from a Magistrate Judge in the same district, which had extended work product protection to communications between a plaintiff and her husband (who was also a lawyer) even without a formal direction of counsel. Judge Rakoff respectfully disagreed with Shih, reinforcing a stricter interpretation of the doctrine. He reasoned that extending protection to materials generated by a client on their own—without the direction of counsel—does not serve the doctrine’s goal of preserving a zone of privacy for the attorney (p. 11).

Because Heppner’s counsel admitted they “did not direct [Heppner] to run Claude searches,” the resulting documents were merely the defendant’s own statements to a third party, leaving them fully discoverable (p. 10).

Key Takeaways and Practical Implications

The Heppner decision provides the first clear judicial guidance on the intersection of generative AI and privilege. While this is a district court opinion, Judge Rakoff is a highly influential jurist, and other courts will likely look to this reasoning. For IP practitioners and business leaders, the decision necessitates immediate adjustments to data handling and litigation strategy.

1. The Terms of Service Are Dispositive

The court’s reliance on Anthropic’s privacy policy underscores a critical technical reality: “public” AI is not private. If a platform’s terms allow for data collection, human review, or training on user inputs, there is no reasonable expectation of confidentiality. In patent litigation, where trade secrets and sensitive technical data are paramount, attorneys must assume that any information entered into a public AI tool (like standard versions of ChatGPT, Claude, or Gemini) creates a permanent, discoverable record that constitutes a third-party waiver.

2. Direction of Counsel is the Dividing Line

The court left open a narrow potential avenue for protection. The opinion repeatedly emphasized that counsel “did not direct” the defendant to use the AI. This suggests that if an attorney directs a client (or an expert) to use a specific, secure AI tool as an agent of the lawyer to generate legal analysis, a work product argument might survive. However, this remains hypothetical. The safest course is for attorneys to perform AI-assisted work themselves or strictly supervise it using enterprise-grade tools with confidentiality agreements that preclude model training.

3. The “Software Argument” Failed

Heppner attempted to argue that using Claude is analogous to using a cloud-based word processor—a tool merely used to record thoughts. The court rejected this comparison. Unlike a passive text editor, generative AI involves an interaction with a distinct entity that processes, retains, and “learns” from the data under its own governance policies. Practitioners should not treat LLMs as passive software; the law treats them as third-party recipients of information.

4. Criminal vs. Civil Discovery

While Heppner is a criminal case, the principles regarding privilege waiver are directly applicable to civil patent disputes. In fact, the risk may be higher in civil litigation where the scope of discovery regarding relevance is broad. Inventors engaging with AI to brainstorm patent claims or refine technical descriptions prior to filing could inadvertently waive privilege or create prior art issues if those conversations are deemed public disclosures to a third party.

How Things Might Have Gone Differently

Judge Rakoff’s opinion invites a counterfactual analysis: could a defendant ever successfully claim privilege over AI interactions? The court’s reasoning suggests two potential pathways, though both remain fraught with risk.

First, the “agency” argument might have traction if the attorney explicitly directs the client to use the AI. The court heavily weighed the fact that Heppner acted on his “own volition” and that counsel “did not direct” the usage (p. 10). If an attorney were to instruct a client to “input this specific dataset into this specific tool to generate a summary for my review,” the argument that the AI is functioning as a sheer instrumentality of the lawyer’s strategy becomes stronger. However, this likely only cures the “work product” defect, not the confidentiality waiver inherent in the platform’s terms of service.

Second, the choice of platform matters. The court focused on Anthropic’s standard privacy policy, which allows for data training and third-party review (p. 6). Had the defendant used an enterprise-grade instance of an LLM with a contractual guarantee of zero data retention and no model training—effectively a “walled garden”—the “expectation of confidentiality” prong might have been satisfied. In that scenario, the AI looks less like a third-party confidant and more like a sophisticated, private calculator.

Other Open Questions

The decision also leaves open the status of specialized “Legal AI” tools. If a law firm provides a client with access to a proprietary legal AI assistant to help organize facts for the case, does that constitute a privileged environment?

While Heppner rules out privilege for public chatbots, it arguably distinguishes them from tools where the “trusting human relationship” (p. 6) is replaced by a trusting contractual relationship between the firm, the client, and the software provider.

Until those specific facts are litigated, however, the Heppner rule serves as the baseline: AI inputs are presumed public.

Conclusion

The Heppner decision reinforces that while technology evolves rapidly, fundamental legal principles remain stubborn. As Judge Rakoff concluded, “AI’s novelty does not mean that its use is not subject to longstanding legal principles” (p. 12).

For the IP community, the message is unambiguous: treat public AI chatbots as you would a stranger in a crowded coffee shop. Anything said to them is likely not confidential, and if that conversation is recorded, it will be available to opposing counsel.

Until specific “legal-grade” AI tools with robust non-disclosure architectures become the standard—and are tested in court—the presumption must be that no privilege attaches to the prompt and chat log.

Disclaimer: This is provided for informational purposes only and does not constitute legal or financial advice. To the extent there are any opinions in this article, they are the author’s alone and do not represent the beliefs of his firm or clients. The strategies expressed are purely speculation based on publicly available information. The information expressed is subject to change at any time and should be checked for completeness, accuracy and current applicability. For advice, consult a suitably licensed attorney and/or patent professional.

Testifying as the head of the agency, Squires provided a comprehensive update on technology deployments, backlog reductions, and statutory interpretation under the current administration.

The testimony offered specific insights for intellectual property professionals tracking administrative priorities and preparing for procedural shifts in patent prosecution and trademark registration.

The Economic Foundation of Intellectual Property

Squires conceptualized the USPTO as the Department of Commerce’s “Central Bank of Innovation” (Opening Statement, ¶ 7). The testimony framed intellectual property units as foundational elements flowing into the broader economy.

Squires observed that “every unit of intellectual property we put into circulation is a potential job, a new business, a competitive advantage, an investible asset, a life-saving drug, all flowing into the real economy” (¶ 7).

This economic characterization signals an administrative focus on the commercialization and monetization of patents and trademarks. Treating intellectual property as a tangible economic driver rather than a purely abstract legal right indicates a policy environment favoring strong patent enforcement.

Squires stated that such innovation leads to “national prowess and global reach” (¶ 7), linking domestic intellectual property policy directly to international economic standing.

Patent attorneys and in-house counsel interpreting this rhetoric can anticipate an administration likely to favor agency rules that support patent validity and enforceability in commercial contexts.

Constitutional Frameworks and Global Positioning

The congressional hearing highlighted a distinct ideological approach, prioritizing domestic interests through an explicit “America First IP” agenda (¶ 14). Squires drew a direct line from the Founders’ constitutional inclusion of intellectual property protection to current global competitiveness. He recounted a recent exchange with the head of the United Kingdom’s Intellectual Property Office regarding the adaptability of the United States framework to emerging technologies.

According to Squires, the United States constitutional framework permits the USPTO to readily adapt to emerging fields such as artificial intelligence and quantum computing (¶ 12), prompting the United Kingdom official to remark, “I know – wish I had that” (¶ 13).

Squires utilized this exchange to contrast the United States with foreign administrative systems, stating, “two democracies, two outcomes, but only one clearly atop the IP world today” (¶ 14).

This posture suggests the agency will aggressively pursue administrative policies that maintain a competitive distance from the European Patent Office and other international bodies, potentially by offering more favorable patentability standards for emerging technologies.

Integration of Artificial Intelligence Systems

A substantial portion of the hearing addressed the integration of artificial intelligence within USPTO operations. Patent practitioners and trademark attorneys should anticipate altered timelines and examination procedures based on the implementation of these specific technologies.

In the trademark division, Squires announced the deployment of an “agentic-AI Trademark Classification tool” (¶ 18). The administrative efficiency gains reported are highly significant. According to the Director, processes that previously required “Five months of manual searching, is now a FIVE-SECOND outcome” (¶ 18).

For trademark attorneys, this suggests a near-immediate processing of classification data, which could drastically compress the timeline between filing a trademark application and receiving an initial procedural review.

For patent prosecution, the agency has integrated an “AI search-assistant” designed to provide patent examiners with “a top ten list of prior art – before the first office action” (¶ 19). Squires asserted this would create “quicker pathways to allowance” (¶ 19). Providing examiners with an automated, curated list of prior art fundamentally alters the traditional search strategy, shifting the initial burden of reference discovery from human manual boolean searches to algorithmic retrieval.

Automated systems are actively identifying fraudulent submissions. Squires reported that “AI fraud detections helped us purge 70,000+ baseless filings – in just under a year” (¶ 20).

The testimony characterized these automated systems not as replacements for human personnel, but as tools intended to “become our Examiner’s super-powers, supplying them with a cadre of agents to deploy as they see fit” (¶ 21).

Backlog Reduction and Administrative Funding

Administrative delays consistently impact applicants and practitioners attempting to secure financing or enforce rights. Squires stated that the agency’s primary administrative focus was “slashing the unacceptable backlog by 50,000” (¶ 15). The Director committed to reducing the backlog by “Another 100,000 to come this year” (¶ 16).

This reduction strategy is tied directly to the agency’s fee-setting authority. Squires expressed confidence that by the end of his tenure, “choking backlogs will be a thing of the past, and improved quality – a confidence-indicator” (¶ 17).

He noted that the agency’s operations rely heavily on the “fee authority so thoughtfully provided by Congress, allowing us to operate like a business, maintain long-term financial stability, and achieve our statutory mission” (¶ 9).

Practitioners should monitor upcoming fee schedules, as aggressive technology deployments and backlog reduction efforts typically require sustained capital expenditure funded by increased applicant fees.

Statutory Interpretation and Policy Adjustments

Squires indicated impending shifts regarding statutory interpretation, specifically referencing the Leahy-Smith America Invents Act. The Director noted that “Congress affords us discretion, and our north star to its exercise is always both the letter and the spirit of the AIA” (¶ 22).

The testimony outlined a dual approach to policy adjustments, focusing on pre-grant examination and post-grant proceedings. Squires stated, “We are restoring balance and fairness on both the front-end – with eligibility determinations – and the back end as to trials and error correction – with new feedback loops in between” (¶ 23).

This explicit mention of “eligibility determinations” signals a deliberate effort to alter how 35 U.S.C. § 101 guidelines are applied by examiners. It suggests the USPTO may issue new guidance directing examiners to apply a more permissive standard for software and diagnostic method patents.

Concurrently, modifications to Patent Trial and Appeal Board procedures appear imminent. Framing PTAB trials as “error correction” mechanisms (¶ 23) often correlates with policies intended to limit the institution of Inter Partes Reviews, providing more certainty to patent owners after issuance.

Thoughts and Considerations

Evaluating the March 25 hearing requires a pragmatic assessment of the USPTO’s trajectory. The aggressive adoption of artificial intelligence within administrative procedures presents specific operational benefits, alongside distinct institutional vulnerabilities that IP professionals must monitor.

The primary benefit lies in operational efficiency and prosecution speed. Reducing trademark classification searches from months to seconds (¶ 18) allows brand owners to secure market positioning much faster. In patent prosecution, if the “AI search-assistant” accurately identifies highly relevant prior art “before the first office action” (¶ 19), patent applicants might experience fewer overall office actions.

Reaching a final disposition quickly reduces prosecution costs and allows inventors to commercialize their assets sooner. Purging over 70,000 fraudulent filings (¶ 20) protects the integrity of the federal register, preventing malicious actors from cluttering the trademark system and delaying legitimate corporate applications.

Adjusting 35 U.S.C. § 101 guidelines on the “front-end” (¶ 23) offers software and biotechnology innovators clearer, more predictable pathways to obtaining protection. If it works.

There are distinct challenges associated with these technological deployments. Integrating “agentic-AI” (¶ 18) into federal agency operations requires rigorous quality control and continuous auditing. Relying on algorithmic tools for prior art searches introduces questions regarding the scope, training data, and diversity of the underlying databases.

If an AI search assistant consistently prioritizes specific types of references or misinterprets complex technical language, the resulting “top ten list of prior art” (¶ 19) might miss highly relevant, yet unconventional, references.

Managing the promised reduction of “Another 100,000” (¶ 16) backlogged applications requires balancing administrative speed with examination thoroughness. Accelerating examination without adequate human oversight risks lowering the presumption of validity associated with issued patents, potentially leading to more litigation later.

The reliance on automated systems poses specific risks for patent practitioners and IP owners. A heavy dependence on AI for fraud detection could result in false positives, inadvertently flagging legitimate applications and forcing applicants to expend resources defending their filings before the agency.

Regarding PTAB procedures, altering “trials and error correction” mechanisms (¶ 23) creates near-term procedural unpredictability. As the USPTO applies its “discretion” under the AIA (¶ 22), previous administrative precedents at the PTAB may lose their predictive value. Stakeholders face the risk of operating within a transitional period where internal guidelines shift rapidly, demanding continuous adjustments from in-house counsel and patent attorneys concerning their post-grant filing strategies.

The Agency’s Trajectory

The March 2026 testimony from Director Squires outlines a technologically aggressive strategy for the USPTO. Framing the agency as a “Central Bank of Innovation” (¶ 7) highlights a purely economic approach to intellectual property administration.

Incorporating artificial intelligence to target administrative backlogs and fraudulent filings demonstrates a commitment to modifying standard examination procedures. The promised efficiency gains offer significant value to applicants seeking rapid commercialization; IP professionals must prepare for the operational shifts these algorithms will introduce to daily prosecution practice.

Close observation of the forthcoming changes to eligibility determinations and post-grant trial procedures remains a primary requirement for accurately advising inventors and corporate stakeholders.

Disclaimer: This is provided for informational purposes only and does not constitute legal or financial advice. To the extent there are any opinions in this article, they are the author’s alone and do not represent the beliefs of his firm or clients. The strategies expressed are purely speculation based on publicly available information. The information expressed is subject to change at any time and should be checked for completeness, accuracy and current applicability. For advice, consult a suitably licensed attorney and/or patent professional.

Patent eligibility remains a notoriously muddy analysis, often frustrating practitioners and courts alike. Yet, the analytical waters appear significantly less muddy when the conclusion seems foregone.

In such instances, examiners and courts frequently deploy what this blog has termed the “Alice Trifecta of Doom”—a predictable rejection sequence used to swiftly invalidate functional claims lacking specific technological implementation.

This ruling illustrates another example of that fatal sequence and reinforces the apparent boundary between visionary, novel concepts and patent-eligible technological advancements.

Background of the Dispute

10Tales, Inc. owns U.S. Patent No. 8,856,030 (the ’030 patent), titled “Method, System and Software for Associating Attributes Within Digital Media Presentations” (Slip Op., p. 2). The patent claims priority to a provisional application filed on April 7, 2003. This early priority date precedes the launch of modern social networking platforms like Facebook and Twitter.

The technology relates to customizing digital media content based on a user’s social network information. The stated goal within the specification is to “attract individuals to content that is personally more relevant and impactful for them as opposed to the user skipping over all or a portion of the message” (p. 2).

10Tales, Inc. v. TikTok

132KB ∙ PDF file

Download

Download

Claim 1, the sole claim at issue, recites a system comprising a server, a computer-readable storage medium, and programming instructions. These instructions direct the system to present a first set of digital media assets to a user, retrieve user social network information from an external source, select a second set of assets based on that information, monitor for a “personalization opportunity,” and perform a “rule based substitution” of the assets (pp. 3-4).

10Tales filed an infringement lawsuit against TikTok Inc., TikTok Pte. Ltd., ByteDance Ltd., and ByteDance Inc. in the United States District Court for the Western District of Texas. The litigation transferred to the United States District Court for the Northern District of California under the supervision of Judge Virginia Kay DeMarchi.

TikTok filed an initial motion to dismiss under Rule 12(b)(6), arguing the claim was directed to patent-ineligible subject matter. The district court denied the motion without prejudice, determining that claim construction was necessary before resolving the eligibility question.

Following claim construction, TikTok moved for judgment on the pleadings under Rule 12(c). The district court granted the motion, concluding that claim 1 was directed to the abstract idea of presenting personalized content and lacked an inventive concept. 10Tales appealed the judgment to the Federal Circuit.

The Court’s Analysis: Unpacking the Central Legal Issue

The central legal issue before the Federal Circuit was whether claim 1 of the ‘030 patent satisfied the subject matter eligibility requirements of 35 U.S.C. § 101. The panel applied the Supreme Court’s two-step framework established in Alice Corp. v. CLS Bank Int’l.

At step one, the court evaluates whether the claim is directed to an abstract idea. The Federal Circuit agreed with the district court, finding the claim abstract. The panel noted that claim 1 recites broad functions: presenting initial assets, retrieving user information, selecting new assets, and substituting them. The court found a critical deficiency in the claim language regarding the actual execution of these steps.

“There is nothing in claim 1 directed to how to implement these steps for modifying digital media content based on a user’s social network information. Claim 1 is instead written to claim ‘only a result’ as opposed to a ‘way of achieving it.’ ... Presenting personalized content to a user based on information about the user ... is an abstract idea.” (pp. 6-7).

At step two, the court searches for an inventive concept sufficient to transform the abstract idea into a patent-eligible application. 10Tales argued that the limitation involving “retrieving user social network information from at least one source external to the presented first composite digital media display” provided this inventive concept (pp. 7-8).

10Tales emphasized the novelty of this feature—e.g., what was conventional in 2003?—pointing out that prior art systems had not envisioned using external social network data to improve digital media content.

The Federal Circuit explicitly rejected the attempt to substitute novelty for subject matter eligibility. The court clarified the boundary between Section 102 and Section 101. The panel stated, “A claim for a new abstract idea is still an abstract idea. The search for a § 101 inventive concept is distinct from demonstrating § 102 novelty” (p. 8). The foresight to use social network data, without specific technical implementation details within the claim, remained an abstract concept in the eyes of the panel.

10Tales presented a second argument at step two, asserting that the limitation requiring a “rule based substitution” of digital media assets made the combination of limitations inventive. The Federal Circuit dismissed this argument by looking back at the procedural history.

During the claim construction phase, 10Tales declined to argue for a specific, narrow construction of the term “rule based” (p. 8). The district court construed the term broadly as “a substitution that happens by application of a rule, rather than on the basis of some discretionary or subjective determination” (p. 8).

The appellate panel concluded that this broad, nonspecific limitation failed to restrict the claimed invention to a particular type of modification, thereby providing no inventive concept.

The Alice Trifecta of Doom

Examiners and courts frequently deploy a predictable rejection framework, often labeled the “Alice Trifecta of Doom.” The sequence begins by boiling the patent claim down to a high-level abstract idea. Next, the evaluator classifies the remaining claim limitations as purely functional. Finally, the analysis dismisses any recited hardware as purely conventional.

A predictable set of patent drafting choices triggers this fatal sequence. The rejection framework frequently activates when a patent specification concedes the integration of “well-known” components or generic hardware.

Written descriptions—especially pre-2014—that include discussions regarding automating a manual process, or excessive emphasis on concepts like improving the user experience, frequently invite this precise recipe for ineligibility conclusions.

The 10Tales decision illustrates this classic Alice chain reaction perfectly. In the ‘030 patent, claim 1 relies on generic recitations such as a “server” and a “computer-readable storage medium” (p. 7).

The patent specification offered no specialized hardware or novel algorithmic structure. Rather than providing technical solutions, as the specification notes, the patent sought to “attract individuals to content that is personally more relevant and impactful for them . . . as opposed to the user skipping over all or a portion of the message” (p. 2).

This reliance on a broad, functional goal, combined with the omission of concrete implementation details, triggered the chain reaction and there was no escape.

The court stripped away the generic server, identified the remaining steps as the abstract concept of customizing information, and categorized the action steps as functional results rather than concrete methods.

Key Takeaways and Practical Implications

The 10Tales (nonprecedential) decision delivers several severe warnings for patent practitioners, inventors, and intellectual property owners operating within the software and digital media sectors.

The primary takeaway involves the absolute necessity of drafting claims that explain the “how” rather than merely claiming the “what.” The “trifecta of doom” remains a constant threat during patent litigation. When patent drafters rely heavily on functional gerunds—such as “identifying,” “creating,” “presenting,” and “retrieving”—without tethering those functions to a specific algorithmic process or specialized data structure, the claims invite invalidation.

Courts will isolate the functional language, classify it as an abstract human activity or mental process, and discard the surrounding generic hardware. To survive Section 101 challenges, the specification must detail the specific technical problem and provide a concrete, step-by-step technological solution.

The decision clarifies the severe limits of relying on a visionary concept without technical scaffolding. The inventor of the ‘030 patent anticipated the value of social media data long before platforms like Facebook dominated the digital ecosystem. That foresight represents a valuable business concept, yet it fails to represent a patentable invention without a specialized technical framework.

The Federal Circuit might have reached a different conclusion had the ‘030 patent claims detailed a specific, unconventional data structure for organizing the retrieved social network information.

Reciting a distinct algorithmic method for mapping the disparate social network data points to the digital media metadata might have provided the necessary inventive concept.

Patent owners cannot rely on the chronological novelty of their ideas. An entirely unprecedented method of organizing data or targeting advertisements will fail under Section 101 if the claims read as a high-level functional wish list.

Practitioners must strictly separate their Section 102 novelty arguments from their Section 101 eligibility arguments. Demonstrating that no one had ever thought to retrieve external data for personalization provides zero defense against an Alice challenge. The CAFC wants to hear the “how.”

The ruling also emphasizes the strategic weight of claim construction in the context of subject matter eligibility. The failure of the “rule based substitution” argument demonstrates a critical litigation trap. Patent owners often seek broad claim constructions to maximize the potential for capturing infringers.

A broad construction can become fatal during a Section 101 analysis. By failing to advocate for a specific, technologically grounded definition of “rule based” during the Markman process, 10Tales left the court with a generalized limitation incapable of supplying an inventive concept.

Litigators must balance the desire for broad infringement reads against the strict requirements of subject matter eligibility early in the litigation process. Securing a narrow, highly technical construction for a key limitation often serves as the only viable defense against a Rule 12(c) motion for invalidity.

Conclusion

The Federal Circuit’s affirmation in 10Tales v. TikTok reinforces the judiciary’s strict policing of software patent boundaries. The decision demonstrates the swift mechanics of the Alice framework applied to functional claims deemed to be lacking specific or structural language. By dismantling the ‘030 patent, the court confirmed that alleged generic computer components combined with novel, yet abstract, business concepts cannot survive scrutiny.

The ruling demands a highly cautious, technically precise approach to intellectual property protection in the software sector. Again, this case is nonprecedential, but the lessons and case law are still worth studying.

Developments in fields relying heavily on data processing and algorithmic targeting are not going anywhere for awhile, but patent applications and their claims clearly require exhaustive technical detail these days.

The abstract idea exception to patentability remains a deadly barrier, requiring practitioners to continuously refine their drafting and litigation strategies. The Federal Circuit’s standard is unfortunately pessimistic: technical solutions require technical descriptions, and claims that smell anything like functional results alone merit zero protection.

The idea of including the “how” is simple, but getting your claims labeled as “technical solutions” and avoiding the Alice Trifecta of Doom seems tougher than ever.

Disclaimer: This is provided for informational purposes only and does not constitute legal or financial advice. To the extent there are any opinions in this article, they are the author’s alone and do not represent the beliefs of his firm or clients. The strategies expressed are purely speculation based on publicly available information. The information expressed is subject to change at any time and should be checked for completeness, accuracy and current applicability. For advice, consult a suitably licensed attorney and/or patent professional.

The joke resonated loudly across the patent bar. It highlighted a severe structural deficiency within the intellectual property system.

Legal professionals acknowledge an uncomfortable reality regarding subject matter eligibility: an absolute truth regarding any specific Alice analysis rarely exists. Subjectivity heavily influences the entire evaluation process.

Until the Supreme Court or Congress steps in, maybe AI tools can help.

The Alice Trifecta of Doom

Examiners frequently deploy a predictable rejection framework, often labeled the “Alice Trifecta of Doom.”

1. The sequence begins by boiling the patent claim down to a high-level abstract idea.

2. Next, the evaluator classifies the remaining claim limitations as purely functional.

3. Finally, the analysis dismisses any recited hardware as purely conventional.

A predictable set of patent drafting choices triggers this fatal sequence.

The rejection framework frequently activates when (A) a patent specification concedes the integration of “well-known” components or generic hardware (“any suitable”).

Written descriptions—especially pre-2014—that include (B) discussions regarding automating a manual process, or (C) excessive emphasis on concepts like “improving user experience,” also frequently invite this precise rejection pattern.

Once activated, recovering the claims requires a seemingly insurmountably high burden of including structural, highly specific, and/or super-narrowing language in the claim.

The Examiner Training Void

The examining corps faces severe operational constraints regarding modern jurisprudence. Interviews with examiners frequently reveal a specific operational deficit.

Examiners require dedicated training regarding the director’s Desjardins decision and MPEP update. Currently, many examiners rely heavily on seemingly outdated USPTO examples generated during prior Alice training initiatives.

Anecdotal reports indicate this reliance stems partly from frustration regarding the limited training hours allocated to the examining corps.

The lack of Squires-era instruction solidifies reliance on the Alice Trifecta of Doom.

A Proposed AI Approach

If the agency deploys a legitimate automated assistant, the system must enforce analytical balance.

Patent practitioners and the agency might evaluate the following structural prompt as a foundation for generating more neutral eligibility assessments.

You are an expert patent attorney specializing in patent subject matter eligibility under 35 U.S.C. § 101 and the Alice/Mayo framework. I will provide you with a patent text below.

Please analyze the subject matter eligibility of the first independent method claim according to the two-step Alice/Mayo framework. In your analysis, actively draw analogies to relevant Federal Circuit pro-eligibility case law (e.g., McRO, BASCOM, DDR Holdings, Contour IP, EcoFactor, Packet Intelligence, Visual Memory) where applicable.

Please structure your response exactly as follows, grounding each point in the specific language of the claims and relevant Federal Circuit precedent:

Step One Analysis (Is the claim directed to a patent-ineligible concept?)

Arguments FOR Eligibility (2-4 points): Explain why the claims are not directed to an abstract idea. Highlight aspects that suggest a specific, concrete technological advance or an improvement to a technological process/machine. Identify similar pro-eligibility cases where you can.

Arguments AGAINST Eligibility (2-4 points): Identify the underlying abstract idea and explain why the claim as a whole is primarily directed to this concept. Identify similar cases that were lacking eligibility if you can.

Step Two Analysis (Is there an inventive concept? - Conduct this step even if you argue for eligibility in Step 1)

Arguments FOR Eligibility (2-4 points): Explain how the claim elements, considered individually and as an ordered combination, amount to “significantly more” than the abstract idea by providing an inventive concept (e.g., unconventional steps, specific technical implementations). Compare claim elements to pro-eligibility cases.

Arguments AGAINST Eligibility (2-4 points): Explain why the additional elements are merely “well-understood, routine, and conventional” activity or generic computer implementation. Compare claim elements to cases that found ineligible claims.

Here is the patent:

[INSERT PATENT CLAIM(S) AND/OR SPEC HERE]

This structural prompt represents a solid starting point for Alice evaluators.

The prompt forces the automated system to construct arguments for eligibility, explicitly requiring the integration of pro-eligibility Federal Circuit case law.

Cases like McRO, BASCOM, and Contour IP provide a necessary counterweight to the high volume of invalidating court decisions.

Requiring a balanced output counteracts the default tendency of evaluators to generate immediate rejections.

Still, AI lies and the prompt is not perfect.

The Emergence of SCOUT

The agency recently accelerated internal testing of a generative artificial intelligence web application named “SCOUT,” an internal program structured for searching, consolidating, and outlining examination materials. Developed within a secure laboratory environment to protect unpublished application data, this platform provides examining staff with access to advanced language models calibrated for specific patent evaluation tasks.

Reports indicate current beta testing focuses on an antecedent verification feature to identify claim inconsistencies under Section 112 alongside a developer assistant for analyzing software code.

The application is anticipated to incorporate a specialized manual search feature, granting examiners immediate access to examination procedures. Agency employees suggest future iterations might assist in drafting the substantive content of office actions, moving the technology beyond basic prior art retrieval.

The structural capabilities of the SCOUT platform present a clear pathway for configuring the system to execute the precise subject matter eligibility analysis mocked in the recent press release.

The fictional “MATTHEW” system relies on celebrity absurdity. The underlying concept of an automated assistant evaluating claims under Section 101 aligns closely with the agency’s actual technological trajectory.

The parody rests uncomfortably close to operational reality, signaling that automated eligibility determinations remain a probable outcome of current testing. Patent practitioners must monitor this deployment closely, as the potential introduction of machine-generated reasoning into formal office actions creates new administrative hurdles for applicants rebutting automated conclusions.

Benefits, Challenges, and Risks

Implementing a balanced prompt structure offers distinct operational advantages. By forcing an automated system to actively search for technological improvements, this requirement disrupts the immediate default to the Trifecta of Doom.

The explicit inclusion of pro-eligibility precedent equips examiners with a broader perspective during prosecution, prompting a more thorough review of the technical specifications.

Developing such a system presents specific engineering hurdles. Instructing an algorithmic model to recognize the highly nuanced technological advances identified in cases like Enfish, BASCOM, or EcoFactor requires precise computational mapping of legal concepts. Without strict boundaries tied to accurate, verified legal databases, a language model might invent case analogies or misinterpret court holdings.

Information security introduces an additional technical obstacle. Any system deployed by the USPTO must operate within a secure sandbox environment to maintain the strict confidentiality of yet-to-be-published patent applications.

Initial testing and calibration of these language models should exclusively utilize public, non-confidential materials to prevent the inadvertent disclosure of sensitive commercial assets.

The primary risk involves an over-reliance on automated outputs. If examiners adopt generated arguments without independent verification, the examination process risks automating and scaling existing biases.

A poorly calibrated system might generate rejections that appear plausible on the surface but lack sound legal reasoning. Such a scenario forces patent applicants to expend significant financial resources rebutting machine-generated logic. Algorithmic tools do not substitute for rigorous human analysis.

The current examination environment, however, presents a difficult irony.

Receiving a boilerplate rejection using the Trifecta of Doom frequently leaves practitioners questioning whether genuine human analysis is presently occurring during manual examination.

Conclusion

The agency’s April Fools parody succeeded by mocking a deeply fractured system. The intellectual property community requires substantive, predictable reform regarding subject matter eligibility. Until structural changes materialize, deploying supervised algorithmic assistants using carefully balanced prompts represents a pragmatic operational strategy.

Developing an autonomous artificial intelligence agent operating without human oversight appears unnecessary. Standard large language models possess extensive familiarity with federal case law published prior to their initial training dates. Supplying these models with specific legal texts, or employing fine-tuned systems trained exclusively on patent jurisprudence, improves analytical accuracy.

The examining corps requires tools that present multiple legal arguments. Examiners must evaluate those arguments and select the most applicable reasoning, relying entirely upon their specialized training and professional experience. This supervised framework likely represents the near-term future of automated assistance at the USPTO and parallel federal agencies. A structured approach protects software and diagnostic innovations from arbitrary administrative evaluation.

Disclaimer: the ideas are solely for experimental use in exploring legal and patent analysis. Such guides, code, prompts, and/or any results are not intended to replace the critical judgment of a qualified professional. It is your responsibility to thoroughly verify all outputs and information as AI models are prone to errors and hallucination. Do not bill clients for time/work performed by AI and/or software tools. Follow all rules in accordance with your state bar and/or ethics and governing body.

This is provided for informational purposes only and does not constitute legal or financial advice. To the extent there are any opinions in this article, they are the author’s alone and do not represent the beliefs of his firm or clients. The strategies expressed are purely speculation based on publicly available information. The information expressed is subject to change at any time and should be checked for completeness, accuracy and current applicability. For advice, consult a suitably licensed attorney and/or patent professional.

The announcement playfully introduced a new agentic system named “McConaughey Agentic Tasking Technology Helping Examiner Workload,” carrying the acronym “MATTHEW.”

The announcement jokes that this system “will help examiners tackle the thorniest of eligibility questions as to whether claims presented are an abstract idea or a patent-eligible invention” (¶ 1). Good luck!

The Proposed System and the Suspension of Precedent

The parody relies on the complete replacement of existing Supreme Court precedent with an automated, celebrity-themed response mechanism. The publication quotes USPTO Director John A. Squires explaining the operational mechanics of the new system. Director Squires stated the system will “greatly enhance our ability to make the close calls—or any call, really—as I herewith also suspend all applicable precedent, including Desjardins, Alice, and Mayo” (¶ 2).

The director offered a replacement standard rooted entirely in the output of the automated tool: “Basically, in terms of eligibility, if MATTHEW says your invention is ‘Alright, Alright, Alright,’ then it’s ‘Alright, Alright, Alright’ with the USPTO” (¶ 2).

The text addresses the structural shift from a two-part inquiry to a three-part output. The director remarked, “Initially, we had some concerns that we would be introducing a three-part test in place of the two-part test under Alice and Mayo, but I think we’ll be al…um, okay” (¶ 3).

The agency claimed to have evaluated other automated solutions before selecting the current iteration. The publication notes a rejected system titled “Binary Eligibility Engaged Translation Language Environment Joint User Interface Computational Evaluator,” or “BEETLEJUICE” (¶ 4). The director noted “coders had some issues in testing when they said the name three times” (¶ 4).

The McConaughey Connection

The selection of Matthew McConaughey for this parody directly references recent, highly publicized intellectual property filings. Matthew McConaughey’s company, J.K. Livin Brands Inc., registered eight trademarks consisting of “video and audio recordings of McConaughey delivering iterations of his ‘alright alright alright’ and ‘just keep livin’‘ lines.” Reports indicate the strategy attempts to utilize federal trademark protection to address the “unlicensed use of AI-generated McConaugheys in ways that current IP law may not cover.”

These specific multimedia filings exemplify the exact type of non-traditional applications organized under the new Category 30 design search codes. Public figures continuously attempt to construct “the widest possible protective moats” around their personas.

Filings under divisions 30.01 and 30.02 will likely increase, which seems to be what Director Squires wants for the USPTO.

High Stakes and the Reception of Satire

The humor embedded in the announcement originates from a genuine, prolonged legal predicament. The satire functions effectively as a comedic exercise. The stakes feel extraordinarily high for patent owners and applicants. The intellectual property community has endured approximately twelve years of the ambiguous two-step framework established by the Supreme Court. Identifying an abstract idea remains highly subjective. The absence of clear statutory definitions forces patent examiners and federal federal judges to rely on analogous comparisons with previous case law. This analogical reasoning frequently yields inconsistent results across different technology sectors.

Replacing a legal framework with a machine outputting a movie quote perfectly encapsulates the frustration felt by patent attorneys attempting to counsel clients on the likelihood of obtaining patent protection. The joke might be received poorly by practitioners actively struggling to secure rights for software developers and medical diagnostics innovators. For these professionals, the pain of arbitrary rejections is entirely real. A parody originating from the examining agency might appear insensitive to the severe financial and strategic losses caused by the current jurisprudential environment.

The acronym’s explicit reference to “Helping Examiner Workload” introduces an additional layer of potential insensitivity directed at the agency’s internal workforce. The Ninth Circuit recently vacated a preliminary injunction in Am. Fed. of Gov’t Employees v. Trump, resulting in the effective decertification of the Patent Office Professional Association (POPA). The appellate ruling permits Executive Order 14,251 to proceed, categorizing the Patents business unit as executing primary national security functions. Consequently, examiners currently lack a formal collective bargaining agreement. This absence leaves a distinct void regarding established protocols for performance metrics, telework allowances, and grievance procedures. Attempting humor regarding examiner workload during a period where employees operate without union representation. Offering more training hours might make the joke sting less.

Patent examiners and practitioners spend countless hours debating the presence of an “inventive concept” instead of, e.g., focusing on unearthing prior art references. The two-step test frequently conflates eligibility with novelty under Section 102.

The agency’s attempt at humor highlights this structural failure. Software patents face exceptionally high invalidation rates in district courts. The examining corps struggles to maintain uniformity across different art units.

Again, this April Fools joke—without offering additional training and credit—falls a little flat.

Future Implications

Patent professionals must consider the potential impact of such high-profile commentary. Does this prompt the Supreme Court to take up a new case to clarify the standard? Does this inspire Congressional action to codify a modern approach to patentable subject matter? Or does this just remind applicants in the software space that eligibility is just a gamble?

The persistent demand for legislative clarity regarding patentable subject matter remains unfulfilled. As the agency proceeds with testing and implementing real automated tools for prior art searching and classification, the intellectual property community must monitor these technological integrations carefully. The tension between administrative efficiency and rigorous legal analysis requires continuous observation.

Software and diagnostic innovators face prolonged uncertainty. Integrating automated systems into formal legal evaluations requires extreme caution.

The patent system relies upon transparent, articulable legal reasoning. Intellectual property professionals will continue advocating for predictability and stability in securing valuable commercial assets.

Getting stakeholders’ hopes up for a second does not seem alright, alright, alright.

Disclaimer: This is provided for informational purposes only and does not constitute legal or financial advice. To the extent there are any opinions in this article, they are the author’s alone and do not represent the beliefs of his firm or clients. The strategies expressed are purely speculation based on publicly available information. The information expressed is subject to change at any time and should be checked for completeness, accuracy and current applicability. For advice, consult a suitably licensed attorney and/or patent professional.

On its face, this notice represents a mandatory administrative check-in under the Paperwork Reduction Act. The agency must periodically ask the public to evaluate the time and financial burdens associated with its standard forms.

While the notice appears routine, the specific changes highlight two immediate regulatory priorities: reaffirming conflict of interest policies as patent examiners transition into private practice and monitoring for potential fraud through misuse of practitioner status.

Examiner Attrition and Conflicts of Interest

The patent profession is observing a notable demographic shift at the USPTO. Due to shifting federal policies, union restructuring, and potential reductions in force, numerous patent examiners are leaving the agency to join industry, private law firms, or in-house corporate teams. This not-so-insignificant exodus seemingly necessitates strict ethical oversight to protect the integrity of the patent prosecution process.

The current information collection introduces a targeted mechanism to address this transition.

“The USPTO is adding an item (Item 6-Undertaking Under 37 CFR 11.10(b)) to this information collection.” (p. 1) This new addition, form PTO-275, directly targets former agency employees.

The regulation at 37 CFR 11.10(b) restricts former USPTO employees from prosecuting patents or representing clients in matters where they had personal and substantial involvement during their government tenure.

Recent enforcement actions highlight the severity of similar ethical breaches, with the Department of Justice securing significant settlements from former examiners. In early 2026, two examiners agreed to pay $500,000 and $122,480, respectively, in civil penalties to resolve allegations of examining patent applications for companies in which they held disqualifying financial interests.

Responding to these specific breaches, USPTO Director John A. Squires issued a memorandum on March 2, 2026, imposing a strict zero-dollar financial threshold for recusal. This new directive eliminates the previous $15,000 de minimis exemption, directing any employee who determines the scope of patent rights to affirmatively recuse themselves from applications where they hold any stock or bonds in the applicant.

By requiring former examiners to formally sign this new form, the Office of Enrollment and Discipline establishes a documented agreement regarding ethical boundaries. The agency uses this signed form to “determine the qualifications of individuals entitled to represent applicants before the USPTO”. (p. 1)

This proactive measure aims to guard against the improper use of insider knowledge and establishes a clear paper trail of professional accountability.

Maintaining the Human Roster

Beyond tracking former examiners, the routine data collection serves as a continuous verification of the active patent bar. Practitioners must regularly submit the “Data Sheet-Register of Patent Attorneys and Agents PTO-107A” to update their status. (p. 1)

The agency leverages this data to “administer and maintain the public roster of attorneys and agents registered to practice before the USPTO, which is accessible through the USPTO website.” (p. 1)

This persistent administrative hurdle provides a baseline defense against algorithmic interference. Recent industry observations note the risk of non-attorneys employing generative artificial intelligence to draft and mass-file patent applications. The statute “permits the USPTO to require information from applicants that shows that they are of good moral character and reputation”. (p. 1)

By demanding human validation through forms like the “Oath or Declaration” (PTO-1209), the agency mandates human accountability for every filing. (p. 1) If an automated tool attempts to submit a document under a fictitious or falsified identity, the absence of that identity on the maintained Register triggers an administrative flag, providing a mechanism for inventors to confirm their representative’s legitimacy.

Likewise, if an individual employs artificial intelligence to abuse the filing process, the Office of Enrollment and Discipline relies on this updated data to locate the responsible party and send necessary letters of investigation or formal sanctions.

This persistent verification process protects against another specific vulnerability: the hijacking of inactive credentials. A bad actor utilizing automated mass-filing tools might attempt to bypass system filters by co-opting the registration number of a retired or deceased practitioner.

By mandating active, human-verified status updates, the agency can flag inactive registration numbers and prevent artificial intelligence from operating under the guise of former attorneys.

Thoughts and Considerations

Practitioners evaluating this updated information collection should weigh the administrative realities of the agency’s requirements.

The formal requirement for former examiners to sign form PTO-275 provides clear ethical boundaries, protecting pending patent applications from potential conflicts of interest.

Furthermore, the methodical identity verification process protects the patent system from unauthorized automated practice. The information collection guarantees that vetted human actors handle sensitive intellectual property matters.

The strict administrative requirements place a measurable burden on active patent attorneys and agents. The agency calculates an “Estimated Total Annual Respondent Hourly Cost Burden: $1,236,849” across the profession. (p. 2) Practitioners face recurring obligations to submit updated data sheets, respond to agency inquiries, and pay associated fees.

For example, the notice lists an estimated 3,293 respondents completing the application for registration, taking 30 minutes each. (p. 2) Submitting a “Reinstatement to the Register” requires a $226 fee. (p. 3)

Former examiners face severe professional consequences if they violate the terms of the new undertaking form.

Additionally, the reliance on public records creates specific vulnerabilities regarding data privacy. The agency states that “All comments submitted in response to this notice are a matter of public record” and warns that “the entire comment-including PII-may be made publicly available at any time.” (p. 3)

Active practitioners communicating with the Office of Enrollment and Discipline must scrutinize their submissions to avoid exposing sensitive personal or client information.

Concluding Thoughts

The revised information collection functions as both a routine paperwork audit and a timely enforcement mechanism. By introducing the formal undertaking for former employees, the agency establishes necessary guardrails for a shifting patent workforce.

Following recent Department of Justice settlements regarding examiner financial conflicts and the USPTO Director’s strict zero-dollar recusal threshold, this documented accountability is a necessary safeguard. Furthermore, maintaining the verified human roster provides an active defense against the unauthorized practice of law by artificial intelligence.

By mandating routine status updates, the Office of Enrollment and Discipline can effectively monitor the Register, ensuring bad actors do not use automated tools to hijack the credentials of retired or deceased practitioners. These administrative measures protect the structural integrity of the patent prosecution process.

Disclaimer: This is provided for informational purposes only and does not constitute legal or financial advice. To the extent there are any opinions in this article, they are the author’s alone and do not represent the beliefs of his firm or clients. The strategies expressed are purely speculation based on publicly available information. The information expressed is subject to change at any time and should be checked for completeness, accuracy and current applicability. For advice, consult a suitably licensed attorney and/or patent professional.

Legal practitioners increasingly utilize generative artificial intelligence applications, such as ChatGPT, Claude, and Perplexity, which “create text, images, or other content in response to user prompts.”

Attorneys deploy these tools for brainstorming, research, drafting, and summarizing complex information. While existing regulations already govern the use of emerging technologies, COPRAC determined that clarifying amendments were necessary. While the risks are pretty obvious, attorneys’ duties and ethics may not be crystal clear.

The committee cited the rapid proliferation of artificial intelligence and “continued examples of fake or ‘hallucinated’ content, including outdated, incomplete, or nonexistent legal authorities appearing in documents filed with the court.” Maintaining privilege and confidentiality is also top of mind.

The integration of both generative and agentic systems offers helpful streamlining capabilities, but attorneys hold a strict obligation to use them consistently with their ethical duties. COPRAC approved the proposed amendments on March 13, 2026, establishing a public comment deadline of May 4, 2026.

This action follows an August 22, 2025, letter from the California Supreme Court. The Court directed the State Bar to consider incorporating principles from its 2023 Practical Guidance for the Use of Generative Artificial Intelligence in the Practice of Law into the formal ethical rules. Furthermore, the Court instructed COPRAC to address “agentic artificial intelligence tools, which can enable systems to autonomously perform tasks or workflows without human prompting.”

Alongside the proposed formal rule changes, COPRAC proposed updates to the 2023 Practical Guidance, which the Board of Trustees will review at its May 2026 meeting (not public yet).

Rule 1.1: Competence and Independent Verification

Competence establishes the baseline requirement for legal representation. The current Rule 1.1 dictates the basic duties of a lawyer regarding skill and preparation. The new proposal incorporates artificial intelligence as a specific example of relevant technology within Comment [1].

A newly drafted Comment [2] mandates that a lawyer “must independently review, verify, and exercise professional judgment regarding any output generated by the technology” (Discussion/proposal).

IP practitioners regularly analyze highly complex engineering documents, chemical structures, and software algorithms. Integrating generative or agentic systems into patent drafting introduces specific verification burdens. An attorney utilizing an autonomous system to, e.g., compare patent claims to a technical specification assumes total responsibility for the accuracy of that mapping.

The independent review requirement prohibits practitioners from relying blindly on machine-generated technical arguments during Office Action responses. Artificial intelligence systems frequently misinterpret subtle differences between an invention and a cited prior art reference.

IP professionals face a strict duty to confirm the technical accuracy and legal sufficiency of every argument presented to the courts or USPTO (or any other authority).

Rule 1.4: Client Communication and Informed Consent

Client communication standards face distinct revisions under the proposal. The newly proposed Comment [5] to Rule 1.4 requires lawyers to initiate conversations with clients regarding technology use.

The disclosure duty triggers when the technology “presents a significant risk or materially affects the scope, cost, manner, or decision-making process of representation.”

In such instances, the lawyer must share “sufficient information regarding the use of technology to permit the client to make informed decisions regarding the representation.”

For IP counsel, this signifies a need to continuously audit exactly which AI-based tools are being deployed. The rule specifies this communication duty persists throughout the duration of the representation. If the tools change, there needs to be a conversation. A single clause in an engagement letters appears to be insufficient.

Lawyers must evaluate several factors, including “the novelty of the technology, risks associated with the use of the technology, scope of representation, and sophistication of the client.”

A solo startup inventor possesses a different sophistication level compared to a multinational technology corporation with a dedicated intellectual property department.

Attorneys representing varied clients must adjust their disclosure strategies appropriately. Law firms might need to draft specific addendums to their engagement letters detailing exactly which commercial platforms the firm utilizes for, e.g., patent searching or brief drafting.

Corporate clients might implement strict guidelines prohibiting certain software and tools, requiring outside counsel to adapt their workflows to maintain compliance.

Rule 1.6: Confidentiality and Data Exposure

Protecting client data remains an absolute priority, especially in intellectual property law. Unfiled patent applications contain highly sensitive trade secrets. Email conversations may contain confidential information. Infringement investigations often contain privileged legal strategies and/or work product. The risk of exposure is high.

The proposed addition of Comment [2] to Rule 1.6 explicitly defines the word “reveal” in the context of digital security. The proposal states that revealing includes “exposing confidential information to technological systems, including AI tools where such exposure creates a material risk that the information may be used in a manner inconsistent with the lawyer’s duty of confidentiality.”

Uploading an invention disclosure document into a public, unsecured web interface likely falls under this definition. The commercial model might retain the inputted text to train future iterations of the software. That retention potentially results in a public disclosure that legally destroys patentability worldwide.

But it is important to note that the “reveal” under these rules would be the initial exposure (e.g., typing, submitting, and/or uploading) and not only when the data is leaked or is seen (e.g., by an unauthorized human). All that is necessary for a “reveal” is a “material risk” now.

Patent attorneys must meticulously evaluate the data retention policies, terms of service, and privacy agreements of any software vendor.

While the rules do not say that “enterprise” models are a true safe harbor, it is clear that private, enclosed enterprise systems that guarantee zero data retention for training purposes present a lower risk profile compared to free, consumer-facing applications.

The burden rests entirely on the legal practitioner to verify the technical architecture of the platform before inputting any client trade secrets.

Working with staff, contractors, and vendors presents similar issues and training must be conducted. For instance, sharing drafts with foreign associates via automated translation tools requires similar scrutiny regarding data privacy and cloud storage security.

The “reveal” is seen as the initial exposure of the data to the high-risk platform and not a later leak or discovery.

Rule 3.3: Candor Toward the Tribunal and Hallucinations

The introduction of fabricated case law into official court records initiated much of the current judicial scrutiny surrounding automation in legal practice. The proposed Comment [3] to Rule 3.3 directly addresses the duty of candor.

The rule states that a lawyer possesses an “obligation to verify the accuracy and existence of cited authorities, including ensuring no cited authority is fabricated, misstated, or taken out of context, before submission to a tribunal.”

The text specifically encompasses “any cited authorities generated or assisted by artificial intelligence or other technological tools.”

Patent litigators draft extensive briefs involving complex claim construction arguments and invalidity contentions. Generative software platforms might produce plausible-sounding but entirely fictitious Patent Trial and Appeal Board decisions or Federal Circuit opinions.

Practitioners submitting briefs to federal courts, the International Trade Commission, or the Patent Trial and Appeal Board carry the absolute burden of reading and verifying every single cited case.

A citation generator might format a citation correctly but invent the volume and page numbers. The proposed rule makes it an ethical violation to submit such a document.

The independent verification of legal citations requires manual checks using traditional, verified legal databases.

Rules 5.1 & 5.3: Managerial Duties and Nonlawyer Supervision

Law firm partners and supervisory attorneys hold ethical responsibility for the actions of their subordinates. The amendment to Rule 5.1 Comment [1] states that “managerial lawyers must make reasonable efforts to establish internal policies and procedures governing the use of AI, in accordance with the Rules of Professional Conduct.”

Law firms cannot ignore the existence of these platforms or rely on unwritten practices. Firms must actively draft, distribute, and enforce formal written usage policies.

Rule 5.3 addresses the supervision of nonlawyer assistants. Paralegals, legal secretaries, and technical specialists frequently assist in preparing Information Disclosure Statements or formatting patent applications.

The proposed modification to Rule 5.3 dictates that a lawyer must provide “appropriate instruction and supervision concerning all ethical aspects of their employment, including the use of technology in the provision of legal services, such as artificial intelligence.”

If a paralegal uses an unapproved automated tool to summarize prior art or translate a foreign patent document, the supervising attorney bears the ethical responsibility. Law firms must implement training programs for all staff members, not merely licensed attorneys.

Big Changes & Takeaways

The “Reveal” Standard and Redefining Confidentiality

California explicitly defines merely exposing confidential information to a technological system as a potential ethical breach. Under the proposed Comment [2] to Rule 1.6, the definition of “reveal” expands to include “exposing confidential information to technological systems, including AI tools where such exposure creates a material risk that the information may be used in a manner inconsistent with the lawyer’s duty of confidentiality.”

The bar for a revealing confidential data has been lowered. For patent practitioners, uploading unfiled invention information to a public platform constitutes a breach, regardless of whether a human outside the firm ever reviews the prompt, files, or output.

The End of “One-and-Done” Client Consent

A standard, boilerplate clause in an initial engagement letter fails to satisfy the proposed requirements under Rule 1.4. The new Comment [5] clarifies that the duty to communicate exists “through the life of the representation based on the facts and circumstances, including the novelty of the technology, risks associated with the use of the technology, scope of representation, and sophistication of the client.”

If a firm introduces a materially new automated workflow mid-litigation, practitioners bear an affirmative duty to secure updated, informed consent from the client.

Banning “Autopilot” Legal Work

The California Supreme Court specifically directed the State Bar to address “agentic artificial intelligence tools, which can enable systems to autonomously perform tasks or workflows without human prompting.”

In response, the proposed Comment [2] to Rule 1.1 mandates that an attorney “must independently review, verify, and exercise professional judgment regarding any output generated by the technology.”

A licensed professional must maintain authority and verify all outputs before final submission.

Strict Supervision and Mandatory Human Training

Failing to train staff on the hazards of emerging technologies carries severe consequences for firm leadership.

Revisions to Rules 5.1 and 5.3 dictate that managerial lawyers “must make reasonable efforts to establish internal policies and procedures” and provide nonlawyer assistants with “appropriate instruction and supervision” regarding technology use.

If a paralegal generates a fictitious case citation using an unapproved tool, the supervising attorney(s) and partner(s) face direct ethical exposure for failing to train the staff adequately.

Ignorance is No Defense

Attorneys cannot claim a lack of technical knowledge regarding fabricated case law to escape liability. The proposed Comment [3] to Rule 3.3 imposes a strict “obligation to verify the accuracy and existence of cited authorities, including ensuring no cited authority is fabricated, misstated, or taken out of context.”

Incompetence regarding a platform’s tendency to hallucinate provides no shield against disciplinary action under Rule 1.1.

Conclusion

The proposed amendments by the State Bar Standing Committee on Professional Responsibility and Conduct reflect a concerted effort to codify ethical boundaries for emerging technologies. The legal profession faces a permanent shift in how daily tasks are executed.

The public comment period remains open until May 4, 2026. Interested parties can submit feedback through the online Public Comment Form, allowing practitioners to influence the final regulatory framework.

Further proposed changes to the existing Practical Guidance will go before the Board of Trustees at its May 2026 meeting.

Monitoring these California developments carries significant weight for practitioners nationwide. California’s proposed framework sets a strict baseline regarding the independent verification of agentic systems and the continuous requirement for informed client consent.

Other major legal jurisdictions frequently observe California’s regulatory approach to emerging technology.

Ethics committees and state bars in New York, Illinois, New Jersey, and Pennsylvania—and potentially the USPTO—will likely study the outcome of this comment period and any approved amendments. Those states possess a high probability of adopting similar, rigorous standards governing autonomous legal workflows.

Legal professionals across the country should review these proposals carefully to prepare for future compliance requirements in their respective jurisdictions.

Disclaimer: This is provided for informational purposes only and does not constitute legal or financial advice. To the extent there are any opinions in this article, they are the author’s alone and do not represent the beliefs of his firm or clients. The strategies expressed are purely speculation based on publicly available information. The information expressed is subject to change at any time and should be checked for completeness, accuracy and current applicability. For advice, consult a suitably licensed attorney and/or patent professional.

The targeted applications typically “have the same or substantially the same disclosure (e.g., domestic parent and counterpart foreign applications) as the U.S. application being examined” (RPA Initiative, ¶ 3).

The extension broadens the program to include one workgroup in each Technology Center, encompassing 61 distinct art units. The Office expects quarterly additions to the program throughout calendar year 2026.

Overview of the RPA Initiative

The core mechanism of the RPA Initiative seeks to leverage existing electronic resources to improve examiner access to relevant information from related applications.

These electronic resources include related U.S. applications, such as continuations and divisional applications claiming benefit under 35 U.S.C. 120 or 121, alongside international applications filed under the Patent Cooperation Treaty, and counterpart foreign applications.

The current Phase 1 focuses exclusively on importing references from an immediate U.S. parent application into a continuing or divisional application. Future phases will incorporate the foreign and international data.

The automated interface aims to equip examiners with a master reference list. This list contains references cited in the instant application alongside imported references from immediate U.S. parent applications. The user interface reportedly allows examiners to filter the reference list and create search strings of U.S. patent documents for use in other search systems.

Historical Context and Program Development

The program originated in fiscal year 2019 following a targeted release in late 2018. The Office gathered user feedback via a Federal Register Notice published in August 2016 and a public roundtable on September 28, 2016.

Initial efforts reportedly received positive feedback from patent applicants and examiners. Resource constraints limited broader implementation until the deployment of a new automated tool.

Research and evaluation preceded the development of this tool. The Office evaluated the potential to import information from existing electronic platforms. The evaluation included Global Dossier, the Common Citation Document, the World Intellectual Property Organization’s PatentScope, and internal information technology systems. The Office developed specific functionality to provide notice to the applicant regarding the relevant information and references imported and considered by the examiner.

Mechanics of Automated Importation

Based on the USPTO’s RPA website and FAQ, the Office determines whether an application meets the conditions for inclusion after the Office of Patent Application Processing completes pre-examination processing. This determination requires the issuance of a filing receipt, the absence of outstanding pre-examination notices, and the completion of classification.

If an application receives a Notice to File Missing Parts for missing fees or an inventor’s oath, the RPA inclusion determination pauses until the applicant remedies the defect.

Once the application meets these conditions, the system executes a single importation of citations from the parent application. The citations encompass U.S. patent documents, foreign patent documents, and non-patent literature contained on an Information Disclosure Statement listing, such as form PTO/SB/08, or a PTO-892 form in the file wrapper record of the parent application.

Applicants receive a Notice of Imported Citations listing the references transferred into the continuing application. The Office specifies that “there is no requirement to respond to the Notice of Imported Citations” (FAQ, ¶ 28). The automated importation occurs behind the scenes. This shifts the initial administrative burden of cross-citing references from the practitioner to the agency’s internal software.

Procedural Requirements for Inclusion

According to the FAQ, Specific criteria must be satisfied for an application to qualify for Phase 1 of the program.

• Application Type: The application must be a non-reissue, non-provisional application filed under 35 U.S.C. 111(a).

• Priority Claim: The application must contain a claim for benefit under 35 U.S.C. 120 or 121 of only a single prior U.S. parent application filed under 35 U.S.C. 111(a) or entered into the national stage under 35 U.S.C. 371. The parent application can claim priority of other applications only under 35 U.S.C. 119.

• Timing: The claim for benefit must be reflected on the filing receipt before the application completes pre-examination processing.

The inclusion criteria restrict the scope of the initial phase. The exclusion of applications claiming priority to multiple U.S. parents prevents the system from managing conflicting citation lists during this early stage.

Reissue applications and provisional applications remain entirely outside the scope of this automated importation. The requirement that the priority claim exist on the filing receipt before pre-examination processing completes forces practitioners to perfect priority claims at the time of initial filing. Correcting a priority claim later in prosecution bypasses the automated importation trigger. This forces the applicant back into manual reporting procedures.

Affected Technology Centers and Art Units

The March 2026 extension includes a diverse array of technical fields. Intellectual property professionals operating in these sectors must adapt docketing procedures accordingly.

• Technology Center 1600 (Biotechnology and Organic Chemistry): Participating art units 1611 through 1619 examine patents covering drugs, bio-affecting compositions, plant-protecting compositions, cosmetics, and in-vivo diagnostic agents. Pharmaceutical patent prosecutors handling large patent families will likely observe immediate impacts on continuing application strategies.

• Technology Center 1700 (Chemical and Materials Engineering): Art units 1731 through 1738 handle applications for coatings, plastics, ceramics, metallurgy, catalysts, and alloys. Patent practitioners practicing in materials science will see automated prior art citations affecting applications in these units.

• Technology Center 2100 (Computer Architecture and Software): Art units 2132, 2133, 2135, 2137, 2138, and 2139 focus on hardware memory, databases, artificial intelligence, error detection, and software development. Software patent practitioners frequently utilize continuing applications to capture evolving commercial embodiments. This makes the inclusion of these units highly relevant for tech-focused portfolios.

• Technology Center 2400 (Computer Networks and Security): Art units 2431 through 2439 process patents involving cybersecurity, encryption, secure network communications, and information security protocols.

• Technology Center 2600 (Communications): Art units 2610 through 2616, 2618, and 2619 are dedicated to computer graphics processing, 3D animation, image analysis, display attributes, and structural modeling. Telecommunications and hardware companies hold significant volumes of pending applications in these units.

• Technology Center 2800 (Semiconductors and Optical Systems): Art units 2871, 2872, 2875, 2876, 2877, and 2878 review liquid crystal cells, optical waveguides, general optical systems, and active solid-state devices.

• Technology Center 2900 (Designs): Art units 2961 through 2964 handle the aesthetic and ornamental design aspects of manufactured products. Design patent practitioners must recognize that the RPA Initiative applies here.

• Technology Center 3600 (Transportation and Construction): Art units 3631 through 3637 review mechanical and structural patents covering supports, racks, joints, connections, and cabinet structures.

• Technology Center 3700 (Mechanical Engineering and Manufacturing): Art units 3752 through 3754 specialize in fluid sprinkling, spraying, diffusing, coating apparatuses, valves, and fire extinguishing systems.

Duty of Disclosure and Legal Implications

The initiative directly interacts with the duty of disclosure under 37 CFR 1.56. The Office states that an applicant’s duty to disclose information in the continuing application “will continue to be satisfied for information considered in the parent application and will be satisfied for any additional information made of record by the Office in the continuing application” (FAQ, ¶ 11).

Compliance with 37 CFR 1.98 in the parent application serves as a strict prerequisite for the examiner’s consideration of the imported documents. If a reference suffered from a procedural defect in the parent application, that defect prevents the automatic consideration of the document in the continuing application. Examples of procedural defects include a missing English translation for a foreign document or an omitted non-patent literature copy. The examiner informs the applicant in the first Office action of the reasons a citation was not considered.

The examiner indicates consideration by signing a Notice of Consideration. All citations imported from the parent application and indicated as considered will be printed on the issuing patent. The Office marks these specific citations with a double-dagger (“‡”) on the patent face to distinguish them from other citations of record.

If an applicant notices one or more citations from the parent application have not been imported, specific actions apply. To have these citations printed on the face of the continuing application, the applicant must submit them on an Information Disclosure Statement. Otherwise, the examiner will still consider the documents corresponding to these citations that have not been imported in accordance with MPEP 609.02.

MPEP 609.02 requires examiners to review the parent application’s file wrapper, but manual submission guarantees printing on the final patent.

Benefits, Challenges, and Risks

The automated importation of prior art offers tangible advantages for patent practitioners and intellectual property owners. The system “Simplifies application process for applicant” and “Speeds up examination process” (RPA Initiative, ¶ 2-3). The availability of a master reference list early in the examination process supports higher quality initial search results and “Improves prosecution quality” (RPA Initiative, ¶ 5).

The current iteration of the program features specific structural limitations requiring careful attention. The Office explicitly states, “In the first phase of the RPA initiative, the Office will perform only a single importation of citations from the patent application” (FAQ, ¶ 19).

Any citations appearing in the parent application after this single importation event will not transfer automatically. Practitioners face the ongoing challenge of tracking parallel prosecution timelines across multiple active matters.

If an examiner cites new art in a parent application after the continuing application has passed pre-examination processing, the applicant must submit a manual Information Disclosure Statement with the later-appearing citations in compliance with 37 CFR 1.97 and 1.98 in the continuing application.

A failure to manually cross-cite later-appearing art from a parent application into a continuing application exposes the resulting patent to allegations of inequitable conduct during future litigation.

Inequitable conduct claims pose a severe threat to patent validity. Courts hold inventors and legal representatives to strict standards of candor. If a practitioner assumes the RPA Initiative imported a highly relevant piece of prior art from a parent case, but a technical glitch prevented the transfer, the resulting patent might face unenforceability claims during enforcement proceedings.

The burden of proof remains on the patent owner to demonstrate compliance with 37 CFR 1.56. The USPTO’s automated tool acts as an administrative aid, not an indemnity against legal malpractice or failure to disclose material information.

Another risk is that automated data transfer systems occasionally experience technical failures. A cautious approach dictates that law firms and corporate legal departments maintain strict internal docketing procedures to audit the Office’s automated importations against internal records.

Complete reliance on an external automated system for legal compliance presents an unacceptable risk profile for sophisticated intellectual property operations. A prudent risk management strategy likely involves treating the Notice of Imported Citations as a secondary verification tool rather than a primary docketing mechanism.

It may be considered best practice for practitioners to systematically compare the Notice of Imported Citations against the parent application’s file history to identify dropped references or unconsidered non-patent literature.

Future Program Phases

The Office plans future phases to focus on importing from additional sources. These sources include counterpart foreign and international applications. Future updates will provide examiners access to text-searchable copies of documents within the master reference list.

The Office anticipates quarterly additions to participating art units throughout the remainder of calendar year 2026. Patent professionals should monitor the RPA Initiative webpage for updates regarding newly participating art units and adjust filing strategies accordingly.

Positioning relevant prior art before examiners earlier in prosecution offers clear advantages. Director Squires frequently points to thorough initial examinations as a basis for denying post-grant administrative procedures.

Issuing valid claims initially becomes more attainable when the most pertinent references from related matters surface immediately.

A lingering question involves how these procedures might alter when the USPTO or counterpart foreign patent offices deploy artificial intelligence tools to identify prior art.

For instance, if Country J employs an automated AI search in its native language immediately upon receipt of an application and payment of fees, prior to the USPTO assigning an examiner, how will these systems categorize the influx of results? The burden might shift heavily onto U.S. examiners to manually review each of those AI-found and potentially machine-translated references.

Such technological adjustments will demand strict audits regarding the relevancy of results, the timing of disclosures, and the accuracy of machine translations. Still, the underlying objective to share the most pertinent references across jurisdictions and quickly locate the best art remains admirable and efficient.

Disclaimer: This is provided for informational purposes only and does not constitute legal or financial advice. To the extent there are any opinions in this article, they are the author’s alone and do not represent the beliefs of his firm or clients. The strategies expressed are purely speculation based on publicly available information. The information expressed is subject to change at any time and should be checked for completeness, accuracy and current applicability. For advice, consult a suitably licensed attorney and/or patent professional.

The software industry historically mandated strict protocols to recreate functionality without infringing on copyrights.

Today, developers use AI agents to execute similar rewrites in a fraction of the time. The practice raises immediate questions regarding derivative works and open-source license enforcement.

Licensing and the Open-Source Ecosystem

Open-source software distribution operates primarily under two licensing categories: permissive and copyleft.

Copyleft represents a legal mechanism used within software licensing to mandate that any derivative works or modifications be distributed under the identical terms as the original software. The mechanism prevents proprietary enclosure of open-source projects, keeping the source code accessible for future modification and distribution.

Permissive agreements take a different approach, allowing developers to incorporate open-source code into proprietary applications with minimal restrictions beyond simple attribution.

The GNU General Public License (GPL) stands as a prominent example of a strong copyleft framework, requiring any combined work to inherit its exact licensing terms.

The GNU Lesser General Public License (LGPL) offers a middle ground, applying copyleft rules to the specific library but permitting proprietary applications to link to it dynamically.

On repositories like GitHub, the MIT License remains exceptionally common, offering a highly permissive structure that limits liability without imposing copyleft obligations.

The Apache License 2.0 provides another popular permissive option, distinguished by an explicit grant of patent rights alongside the copyright license.

The intersection of these varied licensing models with artificial intelligence code generation introduces substantial compliance risk for enterprise engineering teams. Counsel must monitor which licenses attach to the codebases ingested or modified by automated tools to prevent unintended intellectual property contamination.

The Historical Standard for Independent Creation

The technology sector has long relied on specific methodologies to recreate software without violating intellectual property protections.

The standard model requires two separate groups of engineers. The first group analyzes the target software to draft a functional specification. The second group, entirely isolated from the first and from the original source code, uses that specification to write new code.

Simon Willison notes, “The most famous version of this pattern is when Compaq created a clean-room clone of the IBM BIOS back in 1982” (Willison).

This method isolates the ideas, which are not copyrightable, from the expression, which is protected. The process guarantees the new code is not a derivative work. In the past, this required immense financial investment and months of labor.

The Chardet Relicensing Discussion

The theoretical debate surrounding AI code generation materialized recently in an open-source conflict. Mark Pilgrim released the initial version of chardet, a character encoding detector, in 2006 under the GNU Lesser General Public License (LGPL). The LGPL allows proprietary software to link to a library, but mandates that modifications to the library itself be released under the same copyleft terms. Dan Blanchard took over maintenance in 2012.

Blanchard recently released version 7.0.0, describing the update as a “Ground-up, MIT-licensed rewrite of chardet. Same package name, same public API — drop-in replacement for chardet 5.x/6.x. Just way faster and more accurate!” (chardet). The MIT license is permissive, carrying no copyleft restrictions.

The shift from the LGPL to the MIT license triggered an immediate response from Mark Pilgrim.

Pilgrim asserted the maintainers lacked the authority to alter the license. He argued, “doing so is an explicit violation of the LGPL. Licensed code, when modified, must be released under the same LGPL license” (Pilgrim).

Pilgrim rejected the premise of an independent rewrite, stating, “Their claim that it is a ‘complete rewrite’ is irrelevant, since they had ample exposure to the originally licensed code (i.e. this is not a ‘clean room’ implementation). Adding a fancy code generator into the mix does not somehow grant them any additional rights” (Pilgrim).

Algorithmic Measurement Versus Procedural Separation

Blanchard defended the relicensing by focusing on the structural differences in the new codebase. He conceded he possessed deep familiarity with the original code, negating the possibility of a traditional procedural separation. He argued the procedural separation is merely a method to prevent the creation of a derivative work.

Blanchard stated, “It is a means to an end, not the end itself. In this case, I can demonstrate that the end result is the same — the new code is structurally independent of the old code — through direct measurement rather than process guarantees alone” (Blanchard).

To support his position, Blanchard utilized JPlag, a source code plagiarism detection tool. JPlag parses source code into syntactic tokens and evaluates them using Greedy String Tiling, ignoring variable names and formatting. The tool measured an average similarity of 0.50% and a maximum similarity of 0.64% between version 1.1 and the new 7.0.0 release (Blanchard).

Blanchard concluded, “No file in the 7.0.0 codebase structurally resembles any file from any prior release... The MIT license applies to it legitimately” (Blanchard).

The Idea-Expression Dichotomy and AI Prompting

The process used to generate the new code introduces specific legal variables. Blanchard operated in an empty repository without access to the old source tree. He instructed the AI model, Claude, not to base its output on LGPL or GPL-licensed code. He provided requirements including public API compatibility and zero runtime dependencies. Blanchard reviewed and iterated upon the generated code.

The application of copyright law to this workflow remains untested. Courts often apply the Abstraction-Filtration-Comparison test to determine non-literal software infringement. This test separates protectable expression from unprotectable ideas, structural necessities, and elements dictated by external factors.

Blanchard essentially relies on the doctrine of idea-expression dichotomy. He argues the underlying concepts of character detection are “well-established techniques described in publicly available research” (Blanchard). He posits that independently reimplementing these ideas does not create a derivative work.

The Training Data Variable

The integration of large language models complicates the analysis of independent creation.

Willison identifies a significant variable concerning the AI model’s training data. He observes, “Claude itself was very likely trained on chardet as part of its enormous quantity of training data—though we have no way of confirming this for sure” (Willison).

For instance, if the AI model ingested the original LGPL code during its training phase, the output could theoretically carry elements of the original expression. The legal community lacks precedent determining if an AI model acts as an independent engineer or as a tool for automated derivation.

Willison asks, “Can a model trained on a codebase produce a morally or legally defensible clean-room implementation?” (Willison). The answer to this question will dictate the viability of AI-assisted code rewriting.

Future Implications for Software Engineering

The chardet conflict operates as an early indicator of broader industry shifts. Willison anticipates these issues will escalate into commercial disputes. He warns, “Once commercial companies see that their closely held IP is under threat I expect we’ll see some well-funded litigation” (Willison).

The cost reduction in generating functional software from test suites or API specifications alters the economic calculus of software development.

The situation prompts deep questions about AI and intellectual property. Quoting developer Armin Ronacher, Willison highlights the long-term uncertainty: “When the cost of generating code goes down that much, and we can re-implement it from test suites alone, what does that mean for the future of software? Will we see a lot of software re-emerging under more permissive licenses?” (Willison, quoting Armin).

Benefits, Challenges, and Risks

The integration of AI coding agents clearly presents distinct advantages for software modernization, primarily through accelerated performance and resource optimization.

For example, the documentation for the newly generated version of the library notes the software is “44x faster than chardet 6.0.0 with mypyc, 31x faster pure Python” (chardet).

These tools allow developers to upgrade legacy systems at a high velocity, bypassing the substantial financial investments and extended timelines traditionally required by isolated engineering processes.

Automated generation provides the capability to rewrite third-party libraries to eliminate unwanted dependencies, allowing teams to integrate software natively into new environments with fewer resource constraints.

This technological shift introduces significant legal, evidentiary, and technical challenges. The judicial system currently lacks established evidentiary metrics for proving structural independence in AI-generated software.

Tools like JPlag provide quantitative data, yet courts must eventually determine whether syntactic token analysis sufficiently proves non-derivation.

Tracing model inputs adds another layer of difficulty; evaluating the influence of an AI model’s training data presents severe technical hurdles. IP owners face high barriers to proving an AI model replicated protected expression rather than merely adopting functional ideas.

Legal professionals must evaluate whether a developer’s historical exposure to a codebase fundamentally prejudices the prompts provided to the model.

These challenges elevate the broader commercial and legal risks for the intellectual property sector. Widespread automated rewriting threatens the enforceability of copyleft licenses, a foundational element the open-source community relies upon to maintain public access to source code.

In the commercial sector, enterprises utilizing AI to replicate competitors’ proprietary software risk severe copyright infringement liability. The presumption of independent creation weakens when developers rely on models potentially trained on protected data.

Finally, organizations incorporating newly relicensed software face substantial supply chain contamination risks. They may incur legal exposure if courts later rule the AI-generated code constitutes an unauthorized derivative work.

Conclusion

The discussion between the original author and the current maintainer of chardet outlines the emerging tension between traditional copyright doctrines and automated code generation.

Developers possess tools capable of replicating complex software functionality in hours. The legal framework must evaluate whether algorithmic generation directed by informed prompts satisfies the standards for independent creation.

Current copyright doctrines face an unprecedented scaling problem. The judicial system is not currently equipped to handle the sheer volume of disputes that automated code generation will produce. Github isn’t prepared, either.

As developers deploy AI tools to replicate complex software functionality in hours rather than months, the frequency of infringement claims could easily overwhelm existing legal frameworks.

The traditional clean room likely requires structural adaptation. Procedural separation of engineering teams is no longer a practical standard when a single developer can direct an AI model to rewrite an entire library.

The standard for independent creation will likely shift from human isolation protocols to algorithmic verification. It’s unclear if Blanchard’s measurement-based approach should be accepted as the new “clean room” standard.

To manage this transition, the software industry may need to establish automated clearinghouses. Specialized AI review agents could be deployed to audit the outputs of AI code generators.

These clearinghouse models would cross-reference generated code against massive databases of licensed software, verifying that the new code does not inappropriately misappropriate protected expression or violate copyleft licenses prior to deployment.

In the era of vibe coding, the math of a plagiarism detector may eventually carry more weight in court than the testimony of a human developer.

Disclaimer: This is provided for informational purposes only and does not constitute legal or financial advice. To the extent there are any opinions in this article, they are the author’s alone and do not represent the beliefs of his firm or clients. The strategies expressed are purely speculation based on publicly available information. The information expressed is subject to change at any time and should be checked for completeness, accuracy and current applicability. For advice, consult a suitably licensed attorney and/or patent professional.

Addressing this pressing issue, a recent presentation by the National Institute of Standards and Technology (NIST) Information Technology Laboratory (ITL) outlines the agency’s strategic priorities in shaping the international AI standards landscape.

Aligning with the White House’s America’s AI Action Plan, which tasks the Department of Commerce and NIST with dozens of specific directives, the agency is taking a highly proactive stance (slide 3).

Instead of issuing static regulatory mandates, NIST seeks to act as a neutral convener, fostering voluntary consensus standards that can guide both innovators and legal professionals.

Building on a legacy of advancing measurement science, NIST leverages its technical expertise to tackle modern computational challenges while simultaneously acting as the federal coordinator for AI standards across the United States government (slide 29).

The core thesis of this comprehensive initiative is best summarized by the ITL’s stated mission:

“To strengthen trust in AI, accelerate its adoption, and expand U.S. AI dominance by providing the vital measurement science, testing and evaluation, guidance, and standards.” (slide 5)

2026 03 06, Itl Ai Program Webinar, Ai Standards Landscape, Slides, For Web

2.37MB ∙ PDF file

Download

Download

The Problem: Defining “True” Standards Amidst Speed, Scope, and Sociotechnical Complexities

Before addressing the hurdles of AI standardization, it is crucial to clarify what constitutes a “true” standard. The presentation makes a sharp distinction between government guidelines—such as the NIST AI RMF or Cybersecurity Framework—and actual “documentary standards” (slides 13, 14).

As defined by the International Organization for Standardization (ISO) and highlighted by NIST, a true standard is a document “established by consensus and approved by a recognized body” (slide 14). These are largely driven by private-sector Standards Developing Organizations (SDOs) like ISO, IEC, and IEEE. In this ecosystem, the U.S. government is just one stakeholder among many contributing to an industry-led process (slide 26).

Developing these consensus-driven standards for artificial intelligence presents unique hurdles not seen in traditional telecommunications or hardware sectors. First, there is no universally agreed-upon definition of AI; it encompasses a broad spectrum ranging from autonomous agents to generative models, each operating with varying levels of autonomy and adaptiveness after deployment (slide 17).

Furthermore, AI standards often trail the technology itself. Unlike network infrastructures where standards must precede functional hardware so that devices can communicate, AI systems are frequently built and fully deployed within isolated corporate environments long before governance norms or measurement metrics converge (slide 18).

Consequently, these standards focus heavily on sociotechnical factors and risk management frameworks rather than mere functional or mechanical specifications.

The central problem articulated by the presentation is the inherent friction between the velocity of AI advancement and the sluggish nature of international consensus-building.

Developing standards through traditional bodies requires extensive negotiation, which struggles to keep pace with rapid software iteration. The authors capture this duality perfectly:

“AI standards stakeholders have repeatedly emphasized two challenging needs to NIST. We need standards on some of these topics ASAP. We need expertise from a very wide range of stakeholders, including varied types/locations, in AI standards development.” (slide 44)

Ensuring that a broad, inclusive array of voices participates in standardization inherently slows down the deliberative process. However, rushing standards risks codifying brittle definitions that fail to accommodate future technological shifts, ultimately leading to unstable legal frameworks for patent eligibility, bias mitigation, and corporate liability.

The tension between rapid deployment and meticulous consensus is arguably the defining challenge of the current era of technology governance.

Proposed Solution: NIST’s Pre-Standardization and Agile Frameworks

To bridge the gap between technological velocity and rigorous consensus, the NIST presentation proposes a multi-pronged approach focused on pre-standardization research, framework development, and accelerated drafting pipelines (slide 27).

The Zero Drafts Project

“The Zero Drafts project is a pre-standardization effort explicitly meant to feed into formal standards development.” (slide 43)

In practice, the Zero Drafts initiative represents a paradigm shift in how standards are initiated. Instead of waiting for an SDO to begin a project from scratch, NIST is piloting a process where it selects a topic, gathers preliminary community input, and independently authors an advanced “zero draft” (slide 45).

This mature, heavily vetted draft is then submitted directly into the formal SDO pipeline. By front-loading the heavy lifting of drafting and initial consensus-building, NIST accelerates the overall timeline while still subjecting the final document to rigorous international scrutiny.

For legal practitioners, this means key definitions and testing methodologies will reach the market faster, providing quicker clarity for compliance assessments and intellectual property documentation.

Currently, pilot topics include the public documentation of AI datasets and models, as well as a high-level framework for testing, evaluation, verification, and validation (TEVV) (slide 46).

The Artificial Intelligence Risk Management Framework (AI RMF 1.0)

“The AI RMF offers detailed voluntary guidance to operationalize AI governance principles. It has been explicitly referenced in standards.” (slide 36)

It is vital to note that the AI RMF itself is not a standard; rather, it is a foundational guidance document that informs actual standards. The AI RMF moves theoretical AI ethics into actionable corporate practice. It is structured around four core organizational functions: Map, Measure, Manage, and Govern (slide 36).

Rather than acting as a strict regulatory checklist, it provides a flexible architecture for organizations to identify operational contexts, track metrics, and prioritize risks based on projected impacts.

This framework is highly relevant to attorneys advising clients on liability and risk mitigation, as compliance with the AI RMF is increasingly viewed as an industry best practice.

It is already being actively integrated into formal international standards via crosswalks with entities like ISO/IEC and Japan’s AISI, bridging the gap between domestic guidelines and global consensus (slide 37).

AI Agent Standards Initiative

“The AI Agent Standards Initiative ensures that the next generation of AI—agents capable of autonomous actions—is widely adopted with confidence.” (slide 41)

This initiative addresses the emerging frontier of autonomous agentic systems. It focuses on facilitating industry-led standards, fostering community-led protocols, and investing in fundamental research to ensure that AI agents operate securely and interoperate smoothly across digital landscapes.

This is particularly vital for inventors and IP strategists, as interoperability standards will likely dictate the next wave of essential patents and licensing agreements in the AI sector.

By getting ahead of the curve, NIST ensures that the United States maintains a leadership position in defining the boundaries and capabilities of agentic systems before they become ubiquitous in the consumer and enterprise markets.

Examples: Putting Standards into Practice

The presentation highlights several concrete initiatives to demonstrate the breadth of the current standardization ecosystem. These include deep involvement in ISO/IEC JTC 1/SC 42 working groups, the development of crosswalks linking the AI RMF to international guidelines, and the execution of applied measurement challenges (slides 33, 37).

NIST’s direct contributions span dozens of highly specific projects within these committees. For instance, the agency provides crucial expertise to working groups focused on the testing of AI, including red-teaming protocols, as well as separate subcommittees addressing the fundamental security of AI systems (slide 33).

One notable example in the international sphere is ISO/IEC 42001, which provides requirements for bodies providing audit and certification of artificial intelligence management systems (slide 20).

Unlike many standards that merely define terms, 42001 allows an organization’s internal management system to be formally certified by a third-party assessor. This creates a tangible benchmark for corporate compliance, offering a strategic tool for liability reduction and intellectual property due diligence.

Another highly illustrative example of NIST’s pre-standardization research is the GenAI Challenge, an initiative that pits AI content generators against AI “discriminators” or detectors (slide 40). This adversarial research setup is designed to systematically evaluate how well AI-generated content can evade technical detection.

The implications of this research are profoundly important for the legal and intellectual property fields. As the generation of deepfakes, synthetic data, and automated code becomes commoditized, the ability to technically authenticate digital evidence, establish true human inventorship, and protect copyrighted works hinges on the reliable detection methodologies currently being stress-tested in programs like the GenAI Challenge.

By rigorously evaluating these discriminators, NIST is laying the groundwork for what will inevitably become the evidentiary standards of the future.

Conclusion

The effort to standardize artificial intelligence is not merely an academic or bureaucratic exercise; it is the foundational work required to secure and stabilize the global intellectual property ecosystem.

The NIST Information Technology Laboratory’s strategic initiatives, particularly the innovative Zero Drafts project, represent a pragmatic and cautiously optimistic approach to keeping regulatory, legal, and technical frameworks apace with software innovation.

While not every drafted standard will instantly resolve the deep legal ambiguities surrounding AI liability, data privacy, or patentability, the ongoing push toward cohesive measurement science and sociotechnical governance provides a necessary anchor for the industry.

As artificial intelligence transitions from distinct software applications to highly integrated, autonomous agents executing complex workflows, active participation in these standard-setting bodies remains critical. The legal, technical, and inventive communities must continue to engage proactively with these frameworks to ensure that the future of AI innovation is both trusted and legally sound.

This inaugural presentation marks the beginning of a highly promising NIST ITL AI webinar series; the next session, scheduled for April 7, 2026, will tackle the critical technical challenge of building measurement probes into agentic AI ecosystems.

Disclaimer: the ideas are solely for experimental use in exploring legal and patent analysis. Such guides, code, prompts, and/or any results are not intended to replace the critical judgment of a qualified professional. It is your responsibility to thoroughly verify all outputs and information as AI models are prone to errors and hallucination. Do not bill clients for time/work performed by AI and/or software tools. Follow all rules in accordance with your state bar and/or ethics and governing body.

For inventors, intellectual property (IP) professionals, and patent attorneys, this “practical obscurity” has provided a comfortable buffer, allowing employees and researchers to participate in public technical forums without immediately compromising trade secrets or telegraphing corporate initiatives. Generally speaking, your “burner” accounts were relatively safe from being connected or doxxed.

Now, a recent paper for researchers at ETH Zurich titled “Large-scale online deanonymization with LLMs” demonstrates that this buffer is rapidly evaporating.

The research addresses a critical vulnerability in digital privacy, proposing and proving that Large Language Models (LLMs) can automate the extraction and matching of identity signals from raw, unstructured text at scale. As the authors succinctly state, the core thesis of the research is disruptive:

“We show that the practical obscurity that has long protected pseudonymous users (the assumption that deanonymization, while theoretically possible, is too costly to execute broadly) no longer holds.” (Paper, p. 1).

This article examines the mechanisms behind this automated deanonymization pipeline, the empirical results of the study, and the pragmatic implications for professionals tasked with safeguarding proprietary information and navigating litigation in an increasingly transparent digital ecosystem.

Full Citation: Swanson, J., Lermen, S., Aerni, M., Paleka, D., Carlini, N., & Tramèr, F. (2026). Large-scale online deanonymization with LLMs. arXiv preprint arXiv:2602.16800v2.

The Problem: The Collapse of the High-Cost Barrier to Deanonymization

Historically, the ability to uniquely identify individuals from sparse data points is not a novel concept. Researchers have long understood the fragility of anonymized datasets.

However, executing these attacks at a broad scale has always hit a logistical bottleneck. The authors articulate this historical limitation perfectly:

“For decades, it has been known that individuals can be uniquely identified from surprisingly few attributes. Sweeney’s seminal work demonstrated that 87% of the U.S. population could be uniquely identified by just zip code, birth date, and gender [34]. Narayanan and Shmatikov showed that anonymous Netflix ratings could be linked to public IMDb profiles using only a handful of movie preferences [24]...

“Despite these attacks, pseudonymous online accounts (Reddit throwaways, anonymous forums, review profiles, etc) have remained largely unaffected by deanonymization attempts. The reason is simple: applying such attacks in practice has required structured data amenable to algorithmic matching or substantial manual effort by skilled investigators reserved for high-value targets [13].” (p. 1).

This perspective is highly credible and deeply relevant to the fields of cybersecurity and IP management. Security often relies heavily on economic deterrence rather than absolute cryptographic perfection.

If an adversary must spend thousands of dollars in human labor to unmask a single pseudonymous user, widespread dragnet surveillance of forum posts remains economically unviable.

By demonstrating that LLMs can process unstructured text—the very fabric of online communication—with the speed and cost-efficiency of algorithmic matching, the researchers highlight a fundamental shift in the threat model.

The moat protecting casual online discourse has been drained.

Large-scale online deanonymization with LLMs (Backup)

812KB ∙ PDF file

Download

From https://arxiv.org/pdf/2602.16800

Download

Proposed Solution: The Extract-Search-Reason-Calibrate (ESRC) Pipeline

To prove that LLMs can overcome the traditional barriers of unstructured data, the researchers designed a scalable, four-stage attack pipeline.

This methodology does not require bespoke, highly specialized models; rather, it strings together commercially available LLM capabilities to mimic the workflow of a human investigator.

The Extract Stage

“[W]e ask LLMs to identify and structure relevant features from unstructured posts: demographics, writing style, interests, incidental disclosures, etc.” (p. 2).

Instead of relying on neatly organized database columns, this initial step utilizes an AI to read messy, colloquial forum comments and distill them into a structured, easily searchable profile of habits, locations, and technical traits.

The Search Stage

“[W]e encode extracted features into dense embeddings enabling efficient search over thousands or millions of candidate profiles.” (p. 2).

Once a user’s unstructured ramblings are converted into a structured profile, this data is translated into mathematical vectors. This allows the system to rapidly and automatically scan across massive, internet-scale datasets to find candidates with similar attributes, narrowing the field from millions to a manageable shortlist.

The Reason Stage

“[W]e use extended reasoning on top candidates from the search step to identify the most likely match given all available context[.]” (p. 2).

Embedding searches are fast but can lack nuance. In this stage, a more advanced language model acts as a digital detective, reviewing the shortlist of potential matches to cross-reference subtle contextual clues and confirm the identity with a higher degree of accuracy.

The Calibrate Stage

“[W]e prompt LLMs to provide confidences in identified matches (either absolute or relative to other matches), which lets us calibrate the attack to a desired false positive rate.” (p. 2).

A critical feature for any practical intelligence-gathering tool is the ability to self-assess certainty. By forcing the LLM to score its own confidence, operators can filter out weak, speculative guesses and only act on highly probable matches, ensuring the output is actionable and reliable.

Examples: Real-World Testing Across Online Communities

To evaluate the efficacy of the ESRC pipeline, the researchers tested their framework against three distinct datasets featuring known ground-truth data (p. 1).

The experiments included matching users across different Reddit movie discussion communities (p. 1), linking a single user’s Reddit history split across different time periods (p. 1), and linking Hacker News accounts to professional LinkedIn profiles (p. 1).

The Hacker News to LinkedIn experiment is particularly illustrative of the risks posed to corporate and IP ecosystems.

The researchers collected 987 LinkedIn profiles explicitly linked to Hacker News accounts, carefully stripped the Hacker News profiles of all direct identifiers to simulate pseudonymity, and fed them into the pipeline to see if the system could re-establish the connection (p. 6).

The results were striking when compared to classical, non-LLM methods:

“For example, we improve recall from 0.1% to 45.1% at 99% precision when linking Hacker News accounts to LinkedIn profiles[.]” (p. 2).

An improvement from a near-zero success rate to nearly half of all targets accurately identified—at a 99% precision threshold—demonstrates that unstructured professional chatter is highly distinctive.

If an engineer asks a highly specific, anonymized question about a novel polymer on Hacker News, adversaries utilizing this pipeline could trivially link that pseudonymous question back to the engineer’s real-world LinkedIn profile.

The Illusion of the Throwaway Account

The study’s findings severely undermine the presumed safety of “burner,” “throwaway,” or alternative accounts commonly used by professionals to ask sensitive questions without exposing their identity.

By explicitly testing the challenge of “matching a user’s main account to their alt-account” (p. 9), the researchers demonstrated that individuals carry distinctive semantic fingerprints across different platforms and time periods.

Consequently, a highly specific technical inquiry or legal question posted under a newly created pseudonym can still be traced back to the author’s primary public identity by an LLM analyzing their unique traits.

For inventors and attorneys, this confirms that superficially compartmentalizing digital identities is no longer a reliable method for protecting trade secrets or case strategies.

Weaponizing Deanonymization: Litigation, Evidence, and Extortion

The implications of these capabilities extend far beyond corporate espionage; they represent a formidable new frontier in legal strategy and litigation.

The capacity to autonomously match anonymous online histories to verified identities introduces profound risks regarding evidence discovery, witness credibility, and the security of legal counsel.

The paper highlights the broader societal vulnerabilities this creates, noting how threat actors could exploit these systems:

“Hostile groups could identify important employees and decision makers and build online rapport with them to eventually leverage in various forms.” (p. 13).

In a litigation context, this “leverage” takes several concrete forms. First, deanonymization could be weaponized to uncover fraud or undermine testimony.

If an inventor testifies that a specific technology was conceptualized on a certain date, opposing counsel could utilize an LLM pipeline to identify the inventor’s pseudonymous Reddit or GitHub accounts. If those unmasked accounts reveal the inventor seeking troubleshooting advice for that exact technology years prior, the testimony is immediately undermined, and potential fraud is exposed.

Furthermore, the same tools can be directed at the legal system itself. Hackers or opposing entities could target patent attorneys, litigators, clerks, or even judges by unmasking their private, pseudonymous online activities.

The researchers cite related work demonstrating how unmasked data fuels highly targeted attacks:

“[A]dversaries can launch tailored attacks on a user-by-user basis, fundamentally changing the cost-benefit calculus for attackers.” (p. 13).

“[R]ecent work... demonstrate[s] that LLM agents can autonomously crawl public information to construct profiles that were comprehensive for 88% of targets, using them to generate spear phishing emails with click-through rates on par with human experts.” (p. 13).

By uncovering an attorney’s anonymous vents on a legal forum or a judge’s pseudonymous political commentary, malicious actors could formulate sophisticated spear-phishing campaigns to steal confidential case strategies, or worse, use the unmasked activities for direct extortion.

However, a pragmatic view requires acknowledging current limitations within highly sanitized legal documents.

The authors note that applying these models directly to heavily redacted court files remains challenging:

“Nyffenegger et al. [26] evaluate LLM re-identification capabilities on court decisions, finding that despite high re-identification rates on Wikipedia, even the best LLMs struggled with anonymized legal documents.” (p. 12).

While official court redactions may currently present a hurdle, the informal, unstructured text generated by legal professionals and witnesses across the broader internet remains highly vulnerable to automated unmasking.

Closing Thoughts

“Large-scale online deanonymization with LLMs” is a pragmatic wake-up call for data privacy, intellectual property management, and the legal profession. The paper successfully proves that the cost of parsing unstructured text—long considered the ultimate shield for online anonymity—has plummeted.

While the underlying models utilized in the research are not entirely new, the systemic, automated application of these tools to deanonymize users at scale is a sobering development.

For IP attorneys, inventors, and corporate risk officers, this research necessitates a paradigm shift. Trade secrets, competitive intelligence, and litigation strategies can no longer be considered safe simply because related chatter occurs under the guise of an anonymous handle.

As offensive capabilities become cheaper and more automated, the legal and technological industries must look toward funding and developing new defensive protocols—perhaps leveraging the very same LLMs to proactively sanitize employee outputs, audit the digital footprints of key witnesses, or obfuscate identifying semantic markers before they can be weaponized in court.

In the short-term, silence may be the only safe path.

As we like to say around here: Dance like no one is watching. Tweet like one day it will be read in open court.

Disclaimer: This is provided for informational purposes only and does not constitute legal or financial advice. To the extent there are any opinions in this article, they are the author’s alone and do not represent the beliefs of his firm or clients. The strategies expressed are purely speculation based on publicly available information. The information expressed is subject to change at any time and should be checked for completeness, accuracy and current applicability. For advice, consult a suitably licensed attorney and/or patent professional.

The address framed the agency’s identity, explicitly comparing its function to a financial regulatory body. Director Squires asserted that the United States maintains its top global ranking, scoring “roughly 95 percent — a position it has held every year since the Index was first launched” (¶ 8).

The newly released 2026 IP Index provides specific data points corroborating the Director’s concerns, offering patent attorneys, in-house counsel, and inventors additional context regarding international intellectual property enforcement, artificial intelligence implementation, and shifting diplomatic strategies at the World Intellectual Property Organization (WIPO).

An Economic Framework for Intellectual Property

The speech introduced a specific economic framework for the USPTO. Drawing on his background at Goldman Sachs, Squires characterized financial indexes as “decision tools” that reveal “where capital is flowing, where risk is building, and where the next opportunities might emerge” (Address, ¶ 4).

He applied this exact mechanism to innovation policy, stating that “ideas behave a lot like capital” (¶ 4). In his view, capital moves to environments that are “predictable, protected, and capable of supporting long‑term investment” (¶ 4).

For intellectual property professionals, this framing signals a highly macroeconomic approach to agency administration. Squires stated, “That is why we see the USPTO as the Department of Commerce’s Central Bank of Innovation” (¶ 12).

The director detailed that the agency’s mandate mirrors a central bank’s duty to maintain economic stability. Instead of monetary policy, the USPTO manages innovation policy, injecting “soft-dollar intellectual property assets into the stream of commerce” (¶ 17). Every patent and trademark application functions as “the conversion of human ingenuity into a definable, translatable, economic asset class” (¶ 15).

The scale of this economic engine remains massive. The Director noted that IP-intensive industries account for “roughly 41 percent of U.S. economic output” and support “more than 63 million American jobs” (¶ 22). He further stated that IP-intensive services account for about 31 percent of total U.S. services exports (¶ 23).

The Index confirms this assessment, noting that the U.S. services exports surplus contributes “$1.1 trillion to the U.S. economy” (p. 4). Recognizing this scale, the USPTO plans to hire “over 1000 new examiners this year” to fortify the examination process and support increasing filing volumes (¶ 10).

Artificial Intelligence Integration in Examination

Technological progression continues to reshape the examination process. Squires detailed immediate operational updates involving artificial intelligence. Examiners currently utilize AI-enabled platforms “such as Similarity Search, More Like This Document, and image-based design searching” (Address, ¶ 28). Additional capabilities are scheduled for release shortly.

The trademark division has seen aggressive acceleration from these systems. Squires reported that specific classifications and searches that previously required “five months, now takes five minutes. Sometimes, five seconds” (¶ 29).

Beyond artificial intelligence, Squires emphasized the “examination fundamentals of quality and pendency, the basic blocking and tackling side we need to get right” (¶ 29). He defined confidence in an IP system as “inherently and inexorably the three-legged stool of quality, clarity, and timeliness” (¶ 29).

Concurrently, the Index warns that the “rapid expansion of broad AI-related policies created new ambiguities for both rights holders and AI developers” (p. 5).

Intellectual property owners and practitioners generally favor reduced pendency times. Faster initial reviews yield quicker paths to commercialization or clearer signals for pivoting away from unpatentable subject matter.

At the same time, reliance on automated search systems—with less human oversight—warrants caution.

Market Engagement and Standard-Essential Patents

The address outlined the new Standard-Essential Patent (SEP) Working Group. Squires observed that SEPs represent “significant investment, ingenuity, and risk-taking by American inventors as they enable interoperability, create markets, and unlock innovation” (¶ 31).

According to the director, the SEP environment has grown “increasingly hostile amidst widespread efforts to devalue contributions, unclear rules about rights, and systematic suppression of licensing rates” (¶ 32).

The Index provides direct evidence of this trend, reporting that Chinese agencies issued an opinion cementing the government’s role in the SEP licensing process, and the UK launched an SEP consultation that could “introduce uncertainty into SEP licensing markets and devalue the IP of innovative companies” (p. 8).

The creation of the USPTO working group represents a direct administrative intervention into licensing disputes. The stated goal aims to secure “meaningful protection” for American inventors (¶ 32).

For patent holders heavily invested in telecommunications or interoperable software standards, this initiative provides a structured forum to address systemic devaluation efforts.

International Concerns and WIPO Scrutiny

A significant portion of the address focused on global risks. Squires warned that the IP systems of some advanced economies “are in atrophy and decline,” risking the normalization of “lower global standards” (¶ 34).

The Index explicitly corroborates this, reporting that score declines “were largely concentrated in high-income economies traditionally seen as global IP standard-setters, with scores in eight EU Member States declining” (Index, p. 4).

A primary driver involves the EU’s General Pharmaceutical Legislation, which “weakened the framework for IP protection and enforcement” (p. 5) by expanding the Bolar exemption to include commercial activities (p. 6) and shortening the term of regulatory data protection (p. 7).

The USPTO is actively pushing back against this decline through specific enforcement demands. Squires highlighted efforts to “eliminate or narrow overbroad copyright exceptions” abroad (¶ 37). He advocated for “strong civil, criminal, and border enforcement around the world, including satellite and cable signal piracy” (¶ 38). Specifically, the agency supports ex parte authority to allow “the suspension of suspected counterfeit goods without requiring a right holder complaint” (¶ 38).

His remarks directed unusually pointed criticism at WIPO. Squires indicated that WIPO frequently entertains proposals that “weaken intellectual property rights, or attempt to redefine IP rights in ways that undermine the incentives that drive invention” (¶ 39). He stated plainly that the international body’s foundational mission statement “is being forgotten” (¶ 41) and promised that the USPTO “will” remind them of their obligations (¶¶ 42-43).

Squires reminded the audience that WIPO “is the United Nations agency that serves the world’s innovators and creators — ensuring that their ideas travel safely to the market and improve lives everywhere” (¶ 40).

By highlighting this specific mandate, the Director positioned the United States as the true defender of WIPO’s original purpose, contrasting American policy with foreign actors pushing to weaken standards.

For multi-national IP portfolios, this signals an upcoming period of aggressive diplomatic engagement. The USPTO plans to leverage the IP Attaché Program, adding a new post in Seoul, alongside frameworks like IP5, ID5, and TM5 to advocate for American interests abroad (¶¶ 44, 45).

Headwinds in Domestic Pharmaceutical Policy

Yet, IP practitioners should note that the Index identifies specific risks within the United States. The report cautions that U.S. life sciences innovation faces headwinds from “the imposition of Most-Favored Nations drug pricing and proposals to expand march-in rights” (p. 5).

The report flags a new National Institutes of Health policy requiring licensees to submit access plans for medicinal products (p. 8). This context suggests that, alongside international challenges, domestic patent strategies in the biopharmaceutical sector require careful risk assessment regarding government intervention in pricing and licensing.

Thoughts and Considerations

The implementation of artificial intelligence search functions materially reduces wait times for trademark and patent applicants. Accelerated prosecution allows inventors and corporations to secure funding and market positions faster, a goal supported by the agency’s plan to hire over 1000 new examiners.

Recognizing intellectual property as an economic asset class identical to financial capital aligns the USPTO’s mission with broader Department of Commerce objectives. This economic approach validates the heavy financial investments made by in-house counsel and corporate research departments.

Patent owners could benefit from targeted support through the dedicated SEP Working Group, which provides a specialized resource to combat licensing rate suppression in highly standardized technological sectors. Trademark owners receive stronger, immediate defenses at international borders through the agency’s advocacy for ex parte authority against counterfeit goods.

Driving policy changes at WIPO and combating international intellectual property decline requires significant diplomatic leverage. Securing alignment across conflicting jurisdictions remains structurally difficult for multinational filers.

Internally, examining divisions face the task of balancing swift processing times—such as the five-second turnaround in certain trademark classifications—with the statutory mandate for rigorous examination. Practitioners will need to monitor office actions closely to verify that algorithmic speed does not compromise substantive review.

Artificial intelligence models utilized for prior art searches introduce specific vulnerabilities, as they can hallucinate or overlook nuanced technical distinctions. If examiners overly rely on automated similarity features, highly unique or non-obvious inventions might face improper obviousness rejections based on superficial keyword clustering.

On an international scale, the potential fracture between the USPTO and WIPO over fundamental policy directions presents a severe threat. If these entities diverge, intellectual property owners face the prospect of encountering fundamentally incompatible patent enforcement regimes across different continents.

Corporate research investments may migrate to jurisdictions offering the highest protection, leaving multinational filing strategies deeply fractured. For biopharmaceutical innovators, European legislative changes reducing data protection and expanding exemptions present a high risk of eroded market exclusivity.

In the United States, proposed march-in rights and new NIH licensing conditions threaten to restrict the commercial viability of federally funded pharmaceutical inventions.

Conclusion

The March 2026 address establishes a clear, aggressive agenda for the USPTO. By adopting the posture of a Central Bank of Innovation, Director Squires signaled that intellectual property rights are foundational economic assets requiring strong defense against domestic inefficiencies and international dilution.

This outward administrative posture aligns perfectly with broader efforts under Director Squires to promote the agency as explicitly open for business. Recent initiatives reinforce this position, including expanding trademark protections for Name, Image, and Likeness (NIL) rights, creating Category 30 for sound and motion marks to protect against AI fakes, and centralizing Patent Trial and Appeal Board (PTAB) discretion within the Director’s office.

However, the agency’s optimistic public messaging appears incongruent with ongoing executive branch workforce interventions. Recent legal developments, specifically the Ninth Circuit’s ruling permitting the decertification of the Patent Office Professional Association over national security classifications, create significant internal friction. Coupled with proposed Reduction in Force (RIF) rules shifting the agency from a seniority-based system to a metric-driven structure, the USPTO faces severe operational headwinds.

While these new policies aim to strengthen intellectual property frameworks and improve examination quality, the resulting structural instability threatens to deplete the agency’s manpower and institutional knowledge. Patent practitioners and intellectual property owners face a dynamic environment marked by swift technological integration, forceful diplomatic initiatives abroad, and internal administrative volatility.

Continual monitoring of the SEP Working Group, WIPO negotiations, domestic biopharmaceutical policies, and agency staffing levels provides the soundest strategy for anticipating future shifts in enforcement and patentability standards.

Disclaimer: This is provided for informational purposes only and does not constitute legal or financial advice. To the extent there are any opinions in this article, they are the author’s alone and do not represent the beliefs of his firm or clients. The strategies expressed are purely speculation based on publicly available information. The information expressed is subject to change at any time and should be checked for completeness, accuracy and current applicability. For advice, consult a suitably licensed attorney and/or patent professional.