USPTO AI Practitioner Guidance

What is the USPTO AI practitioner guidance?

On April 11, 2024, the USPTO published “Guidance on Use of Artificial Intelligence-Based Tools in Practice Before the United States Patent and Trademark Office” at 89 Fed. Reg. 25609. It is the first federal-agency practitioner-facing AI guidance in the United States. It tells registered patent and trademark practitioners what they may and may not do when using AI tools in submissions to the office.

No new rules are created. The guidance is interpretive, not substantive rulemaking, and has no independent force of law. It applies the existing USPTO rules of practice (37 CFR Part 11), the candor and good-faith duties owed to the office (37 CFR 11.18 and 1.56), and the signature and certification requirements (37 CFR 1.4) to AI-tool-assisted work product. Practitioners are bound by the underlying rules. Failure to follow them, whether or not the guidance was observed, can draw discipline from the USPTO Office of Enrollment and Discipline.

Practitioners not before the USPTO have two reasons to read it: it sets the first articulated federal-agency standard, and it is the template other federal agencies are most likely to copy. Practitioners before the SEC, DOJ, EOIR, and similar federal forums should treat it as a forecast of what their agency will publish next.

Who must comply?

The guidance interprets duties that bind:

• Registered patent attorneys and patent agents practicing before the USPTO under 37 CFR 11.
• Trademark practitioners representing applicants in trademark matters before the USPTO.
• Pro se applicants and party representatives in USPTO proceedings, to the extent 37 CFR 11.18 reaches all submissions.

Patent prosecution carries the sharpest exposure. Patent submissions involve frequent technical drafting, which is where AI assistance shows up most often. Trademark practitioners face narrower exposure within the same compliance perimeter.

What does the guidance require?

Three duties get reaffirmed as applied to AI use:

Duty of candor (37 CFR 11.18, 1.56). Every signed paper certifies that the practitioner has made an inquiry reasonable under the circumstances and that statements are true to the practitioner’s knowledge, information, and belief. AI-generated content gets no exemption. The practitioner stays responsible for verifying every assertion in a paper, however it was drafted. Submitting unverified AI-generated assertions violates the duty whether or not the AI use was disclosed.

Duty of signature (37 CFR 1.4). Each paper must be signed by a person authorized to sign. AI tools cannot satisfy the signature requirement. Applying an AI-generated signature, or arranging for an AI tool to apply the practitioner’s signature, violates the rule. The signature is the practitioner’s own certification.

Duty of confidentiality. Submitting client or invention information to a third-party AI tool may amount to disclosure to the provider, depending on the provider’s data-handling practices. The guidance recommends due diligence on the tool’s data-handling, retention, and training-data practices before any confidential information goes in. The analysis parallels ABA Formal Opinion 512 on Model Rule 1.6.

What does compliance documentation look like?

A registered USPTO practitioner using AI tools should keep:

• A verification protocol requiring every assertion of fact, citation, or technical claim in a USPTO submission to be independently verified before signature.
• A per-submission record noting whether AI tools were used, for what purposes, and what verification steps were performed.
• A vendor-due-diligence record for each tool the practitioner uses, addressing data handling, prompt retention, and training-data exposure.
• A firm-level AI policy the practitioner can produce on inquiry from the Office of Enrollment and Discipline.

The verification protocol carries the weight. Reported failure modes in patent prosecution are fabricated prior-art citations and fabricated technical specifications, and the office has surfaced both during examination rather than during practitioner review.

How does this interact with state bar rules?

A patent practitioner licensed in a US state owes duties to both the USPTO and the state bar. State Rules of Professional Conduct (typically 1.1, 1.6, 3.3, 5.1, 5.3) govern the state-bar dimension. The USPTO guidance governs the USPTO dimension. The two are usually congruent. State-RPC verification, candor, and confidentiality duties are not weakened by the guidance, and in some respects are stricter.

For state-by-state attorney AI guidance, see the state bar guidance tracker.

Primary sources

• Guidance on Use of Artificial Intelligence-Based Tools in Practice Before the United States Patent and Trademark Office, 89 Fed. Reg. 25609 (April 11, 2024): Federal Register entry.