What Published USPTO Guidance, Federal Circuit Opinions, and PTAB Decisions Through October 2026 Actually Say About AI-Assisted Patent Prosecution, Inventorship Disclosure Obligations, and the Duty of Candor When AI Tools Generate Claim Language

The Framework in Brief

Patent prosecutors are operating under a layered framework that emerged in stages between 2022 and 2026: a Federal Circuit decision that resolved the threshold human-inventorship question, two rounds of USPTO formal guidance addressing inventorship and candor respectively, a stream of subsequent Director and examination-unit communications, and a duty-of-disclosure rule that was already on the books before generative AI became a prosecution tool. Understanding what each layer actually requires — and distinguishing it from what commentators have sometimes assumed it requires — is essential for any IP team that uses AI tools in claim drafting, prior art analysis, or prosecution strategy.

The Federal Circuit Baseline: Thaler v. Vidal

The Federal Circuit's 2022 decision in Thaler v. Vidal, 43 F.4th 1207 (Fed. Cir. 2022), is the indispensable starting point. The court held that under the Patent Act, an "inventor" must be a natural person. The plaintiff had sought to name an AI system — the DABUS machine — as a sole inventor on two applications. The court refused, reasoning that the Patent Act's consistent use of personal pronouns and the phrase "whoever invents" reflects a congressional intent to confine inventorship to human beings. The court expressly declined to opine on whether and how human use of AI tools might satisfy inventorship requirements, calling that a question for another day.

That question — the "another day" problem — is the one the USPTO has since been working to answer through administrative guidance rather than through rulemaking. This matters procedurally: USPTO guidance issued under the Director's authority does not carry the force of a legislative rule, but it does bind examiners in their day-to-day work and signals the agency's litigation and examination posture.

USPTO Inventorship Guidance: February 2024

The USPTO published formal inventorship guidance in February 2024, specifically addressing the question of when a human who uses an AI system qualifies as an inventor on the resulting claims. That guidance applied the Pannu v. Iolab Corp., 155 F.3d 1344 (Fed. Cir. 1998) significant-contribution factors to AI-assisted inventions. That February 2024 guidance was, however, rescinded in its entirety and replaced by the USPTO's Revised Inventorship Guidance for AI-Assisted Inventions published November 28, 2025 (90 FR 54636), which withdrew the Pannu-factors approach as inapplicable where a single natural person uses an AI tool, while reaffirming that the traditional conception-based inventorship standard governs.

The November 2025 revised guidance, which superseded the February 2024 guidance, focuses inventorship on the traditional conception standard: a human must have formed a "definite and permanent idea of the complete and operative invention." Where a human prompt to an AI system is itself sufficiently specific and technically substantive to reflect genuine conception, that human can qualify as an inventor even if the AI generated claim language or structural detail. Where a human simply accepts, without meaningful intellectual contribution, an AI-generated output that constitutes the inventive concept, that human cannot claim inventorship solely on the basis of having operated the tool or reduced the concept to practice.

The November 2025 revised guidance confirms that when multiple natural persons collaborate with an AI system, traditional joint-inventorship principles — including the Pannu factors and the standard Ethicon, Inc. v. United States Surgical Corp., 135 F.3d 1456 (Fed. Cir. 1998) established for joint inventorship — continue to apply. The AI system itself cannot be named as a co-inventor under Thaler.

Practical upshot for prosecutors: the inventorship analysis is claim-specific and contribution-specific. A practitioner who uses an AI tool to generate a dependent claim's specific structural limitation must ask whether any named inventor contributed significantly to the conception of that limitation — not merely whether a human "used" the tool.

USPTO Guidance on AI and the Duty of Disclosure: April 2024

The April 2024 USPTO guidance, published in the Federal Register on April 11, 2024 (89 FR 25609), addressed a distinct but related question: what obligations the existing duty of candor and disclosure imposes on practitioners who use AI tools in prosecution. The guidance expressly does not prohibit AI-assisted drafting. It does not require practitioners to disclose to the USPTO, as a threshold matter, that AI was used in preparing a filing.

What the guidance does require is substantive rather than procedural. Under the existing framework of 37 C.F.R. § 11.18 (signature and certification rules) and 37 C.F.R. § 1.56 (duty of disclosure), every registered practitioner who signs a filing takes responsibility for its accuracy and completeness. The April 2024 guidance makes explicit that this responsibility applies equally to AI-generated content: a practitioner who submits claims, specifications, or prior art searches generated by an AI tool cannot rely on the tool's output as a shield against candor obligations. The practitioner must independently verify the accuracy and completeness of AI-generated content before signing and submitting it.

This verification requirement has direct implications for citation accuracy. The Mata v. Avianca problem — ChatGPT-hallucinated case citations submitted to a court — has a patent prosecution analogue: AI tools can generate prior art characterizations, claim charts, or specification language that contains technical inaccuracies or misrepresents prior art. The April 2024 guidance places squarely on the practitioner the burden of catching those errors before submission.

The guidance does not create a new disclosure category for AI use itself. There is no checkbox, no cover-sheet field, and no IDS line item for "AI was used in preparing this application." The obligation runs through the existing candor and competence framework.

Subsequent Director Communications and Examination Guidelines

Following the two formal guidance documents, the USPTO Director's Office issued supplementary examination guidance addressing how examiners should handle prosecution files in which AI-assisted claim generation is apparent or suspected. Consistent with the February and April 2024 guidance, examiners have been instructed that the mere use of AI tools does not, standing alone, give rise to a rejection or inquiry. Examiners are not directed to issue office actions requiring practitioners to certify or disclose AI use as a general matter.

Where an examiner has reason to believe that a named inventor may not have made a significant contribution to the conception of a claimed invention — based on the record as a whole, including declaration content and interview representations — the existing tools of obviousness, anticipation, and inventorship challenge remain available. The guidance does not create a new examination pathway for "AI-assisted prosecution" as a distinct category.

PTAB and the Candor Intersection

Published PTAB decisions through October 2026 have not developed a distinct body of law specifically addressing AI-generated claim language as a standalone category. PTAB proceedings that have implicated AI-assisted prosecution have generally done so through existing inventorship and obviousness frameworks. Where a party in an inter partes review raises questions about inventorship based on alleged AI contribution to the claimed conception, the analysis maps directly onto the Pannu significant-contribution standard as elaborated in the February 2024 guidance.

The candor dimension is more likely to surface in the context of inequitable conduct allegations in district court litigation than in PTAB trial proceedings, given that inequitable conduct is an equitable defense to infringement not available as a ground for IPR or PGR. Practitioners should anticipate that undisclosed AI generation of claim language that materially misrepresents the inventive contribution — particularly where it results in incorrect inventorship — could form the factual predicate for an inequitable conduct defense under the Therasense, Inc. v. Becton, Dickinson & Co., 649 F.3d 1276 (Fed. Cir. 2011) standard, which requires both but-for materiality and specific intent to deceive.

What the Framework Actually Requires: A Practitioner Summary

The current state of the law can be distilled to four operational obligations:

Inventorship accuracy: Each named inventor must have made a significant contribution to the conception of at least one claim, applying the traditional conception standard as confirmed by the USPTO's November 2025 revised guidance (which superseded the February 2024 guidance). AI tools can assist conception without displacing human inventorship, but humans who merely accept AI output without substantive intellectual contribution do not satisfy the standard.

Verification before signing: Under 37 C.F.R. § 11.18 and § 1.56, the signing practitioner is responsible for the accuracy and completeness of all submitted content, including AI-generated claim language, specifications, and prior art characterizations. The April 2024 guidance makes this explicit. There is no AI-output exception to the certification obligation.

No affirmative AI-use disclosure requirement: Neither the February nor the April 2024 guidance requires practitioners to disclose, as a formal filing matter, that AI tools were used in prosecution. The obligation is substantive accuracy and correct inventorship, not process transparency.

Inequitable conduct exposure: Where undisclosed AI contribution results in incorrect inventorship or materially inaccurate representations to the Office, the Therasense inequitable conduct framework applies. Practitioners should document the inventorship analysis contemporaneously, particularly where AI tools contributed meaningfully to claim language or specification content.

The guidance framework, taken as a whole, integrates AI-assisted prosecution into the existing duty-of-candor architecture rather than creating a parallel disclosure regime. That integration places the analytical weight squarely on practitioners to understand what their AI tools are doing and to take genuine intellectual ownership of what they submit.