AI tools are now embedded in ordinary R&D workflows. Engineers and scientists use them to screen compounds, generate code, test design alternatives, model physical systems, draft technical descriptions, and explore solution paths. That use raises a recurring patent question: when AI assists an inventor in conceiving an idea or reducing it to practice, must the applicant tell the patent office?

The current answer in both U.S. and European practice is that routine AI assistance, by itself, creates no stand-alone obligation to disclose AI use in a patent application. The legal focus remains on correct inventorship, accurate patent-office submissions, and sufficient technical disclosure. The controlling inquiry is whether a legally recognized human inventor conceived the claimed invention, and whether any AI-related fact is material to patentability or to the truthfulness of the applicant’s statements.

United States: From the 2024 Guidance to the 2025 Revision

The USPTO’s February 2024 guidance was the agency’s first detailed framework for AI-assisted inventorship. It confirmed that AI-assisted inventions could be patentable where a natural person made a significant contribution to the claimed invention. The 2024 guidance applied the Pannu joint-inventorship factors to evaluate whether the human contribution was sufficient, including in single-human-plus-AI scenarios. Inventorship Guidance for AI-Assisted Inventions, 89 Fed. Reg. 10,043, 10,047–48 (Feb. 13, 2024); Pannu v. Iolab Corp., 155 F.3d 1344, 1351 (Fed. Cir. 1998).

The USPTO revised that approach in November 2025. The 2025 guidance rescinds the 2024 inventorship guidance and withdraws use of the Pannu framework for AI-assisted inventions as a general analytical tool. The Office explains that Pannu applies when multiple natural persons may qualify as joint inventors. When one natural person uses AI, the inquiry is whether that person conceived the claimed invention under traditional conception principles. Revised Inventorship Guidance for AI-Assisted Inventions, 90 Fed. Reg. 54,636, 54,636–37 (Nov. 28, 2025).

The 2025 guidance brings the analysis back to black-letter inventorship doctrine. Under Thaler v. Vidal, AI cannot be named as an inventor because the Patent Act requires a natural person. Thaler v. Vidal, 43 F.4th 1207, 1212 (Fed. Cir. 2022). Conception remains the touchstone: the inventor must have a definite and permanent idea of the complete and operative invention. Burroughs Wellcome Co. v. Barr Labs., Inc., 40 F.3d 1223, 1228 (Fed. Cir. 1994). The USPTO now describes AI systems as instruments used by human inventors, comparable to laboratory equipment, software, or research databases. Revised Inventorship Guidance, 90 Fed. Reg. at 54,637.

For U.S. filings, there is generally no need to include a statement that AI assisted conception, experimentation, simulation, or reduction to practice. The USPTO presumes that the inventors listed in the application data sheet or oath/declaration are the actual human inventors. An application listing an AI system or other non-natural person as inventor should receive a rejection under 35 U.S.C. §§ 101 and 115. Revised Inventorship Guidance, 90 Fed. Reg. at 54,637.

The disclosure issue arises when AI involvement affects inventorship or patentability. The 2024 guidance stated that the USPTO was leaving the duty of disclosure unchanged, while reminding applicants that information showing improper inventorship can be material under 37 C.F.R. § 1.56. The guidance gave the example of evidence that a named inventor’s purported contribution was made by an AI system. Inventorship Guidance, 89 Fed. Reg. at 10,050.

The USPTO’s April 2024 AI-tools guidance reaches the same practical result for patent drafting and prosecution. It states that there is no per se requirement to notify the USPTO when AI tools are used in invention creation or USPTO practice. Where AI use is material to patentability under 37 C.F.R. § 1.56(b), the information must be disclosed. The Office also warned that relying on AI output without human review fails the reasonable-inquiry obligation under 37 C.F.R. § 11.18(b). Guidance on Use of Artificial Intelligence-Based Tools in Practice Before the United States Patent and Trademark Office, 89 Fed. Reg. 25,609, 25,612–13 (Apr. 11, 2024).

The 2025 guidance also adds an important priority point. A U.S. priority claim to a foreign application naming an AI tool as the sole inventor will not be accepted. Where a foreign application names both a natural person and a non-natural person, the U.S. application data sheet should identify only the natural person or persons, with at least one natural-person inventor in common. Revised Inventorship Guidance, 90 Fed. Reg. at 54,637.

Europe: EPO Practice and Recent DABUS Developments

European patent practice should be analyzed through the European Patent Convention and EPO procedure. The EPO Guidelines state that every application must designate an inventor, and that the inventor must be a natural person. European Patent Office, Guidelines for Examination in the EPO, pt. A-III, § 5.1 (2025).

The key EPO decisions remain the DABUS cases. In J 0008/20 (Designation of Inventor/DABUS), ECLI: EP:BA:2021:J000820.20211221 (E.P.O. Legal Bd. App. Dec. 21, 2021), the Legal Board of Appeal dismissed the appeal and stated the catchword: “A machine is not an inventor within the meaning of the EPC.”

The more recent EPO decision, T 0528/25 (Designation of Inventor/DABUS), ECLI:EP:BA:2026:T052825.20260205 (E.P.O. Tech. Bd. App. Feb. 5, 2026), is especially useful for prosecution strategy. The Board confirmed that, under Article 81 EPC and Rule 19 EPC, the inventor designation is a formal assessment, and the EPO does not verify the accuracy of the designation under Rule 19(2). The designation still must identify an inventor within the meaning of the EPC, and the designated inventor must be a person with legal capacity.

T 0528/25 also confirms that the EPC does not exclude patent protection merely because AI was used in developing the invention. The Board quoted J 0008/20 for the proposition that how the invention was made “apparently plays no role” in the European patent system, and stated that inventor designation is possible in principle for inventions developed using AI. T 0528/25, ECLI:EP:BA:2026: T052825.20260205, reasons 1.5–1.6.

The same decision also shows the risk of voluntary AI-attribution language. Dr. Thaler’s EPO Form 1002 named him as inventor, while the addendum stated that DABUS conceived the invention and that Dr. Thaler was not an inventor under traditional criteria. The Board found the statements unclear and contradictory under Article 81 EPC and Rule 19 EPC. T 0528/25, reasons 2.1–3.3.

The Board discussed two national developments that in-house teams should track. It summarized the German Federal Court of Justice’s decision, X ZB 5/22 (June 11, 2024), as allowing designation of a natural person where AI was used to find the technical teaching and where a human contribution significantly influenced the overall success. It also summarized the Swiss Federal Administrative Court’s decision, B-2532/2024 (June 26, 2025), holding that AI cannot be an inventor and that the inventor must be a natural person, where the person exercised extensive control over the AI (data, training, acceptance of outputs, recognition of the invention); Switzerland thereby aligns with the EPO, US, and UK in rejecting AI as inventor while allowing patents that list qualifying human inventors T 0528/25, reasons 1.7–1.8.

The United Kingdom remains relevant for comparative strategy. In Thaler v. Comptroller-General of Patents, Designs & Trade Marks, [2023] UKSC 49, the UK Supreme Court considered whether UK patent law required a person to be named as inventor where the applicant asserted that the invention was created by DABUS without a traditional human inventor. The UKIPO had refused the applications, and the Supreme Court affirmed that refusal.

Is There a Duty to Mention AI Use?

In the United States, routine AI assistance does not require a statement to the USPTO. Counsel should analyze whether the named human inventors conceived the claimed subject matter and whether any AI-generated embodiment, example, claim, or technical assertion creates a material issue under Rule 56 or a reasonable-inquiry issue under § 11.18. Where AI generated claim-relevant material that no named human conceived, disclosure, claim amendment, or claim cancellation may be required. Inventorship Guidance, 89 Fed. Reg. at 10,050; Guidance on Use of AI-Based Tools, 89 Fed. Reg. at 25,612–13.

In EPO practice, there is no general requirement to disclose that AI assisted invention development. The application must contain a clear natural-person inventor designation. Voluntary statements describing AI as having “conceived” the invention can create formal problems where those statements conflict with the human inventor designation. T 0528/25, reasons 2.1–3.3.

Practical Advice for In-House Counsel

Invention disclosure forms should include an AI-use question. That field should be framed as an internal diligence tool rather than a patent-office disclosure requirement. A practical prompt might ask whether AI tools were used in ideation, development, testing, reduction to practice, data analysis, or drafting of the technical disclosure. Follow-up fields should ask for the tool used, the AI outputs relied upon, the human modifications or selections, and the claim features affected.

This practice is supported by the current commentary. AIPLA agreed that existing disclosure and reasonable-inquiry duties are sufficient and that no new AI-specific disclosure requirement is necessary. AIPLA, Comments on Inventorship Guidance for AI-Assisted Inventions (May 15, 2024). Lowenstein Sandler recommends that companies consider updating invention disclosure forms to ask whether inventive features were obtained using AI and whether a human significantly contributed to those features. Lowenstein Sandler LLP, USPTO Issues Inventorship Guidance for AI-Assisted Inventions—AI-Assisted Inventions Are Not Categorically Unpatentable (Feb. 27, 2024). IPWatchdog commentary likewise recommends updating invention disclosure protocols and preserving significant AI inputs, outputs, models, and datasets where they may later evidence human contribution. David S. Kim & Steven N. Levitt, Digging Into the USPTO’s AI Guidance: Adjusting Practices to Capture Human Contribution, IPWatchdog (Feb. 22, 2024). Your prior disclosure-form paper reaches the same practical conclusion: asking about AI use is a best practice for inventorship diligence, prosecution planning, and litigation readiness.

For prosecution, counsel should conduct a claim-level inventorship review whenever AI use appears in the invention record. The review should identify who conceived each claim feature, who selected or modified any AI output, who recognized the claimed solution, and which experiments or simulations confirmed it. For litigation readiness, preserve significant AI interactions in a controlled repository. Full archiving of every immaterial prompt may be unnecessary, although prompts and outputs that influenced claim language, embodiments, examples, or inventor identification should be retained.

For European filings, keep inventor designations clean. Optional statements crediting an AI system with “conception” can undermine a human designation. Where an AI tool must be mentioned in the specification for technical reasons, describe it as a computational tool used by the human inventor or R&D team, and avoid language suggesting that the AI system is the legal inventor.

Conclusion

The U.S. and European positions are now broadly aligned. AI may assist conception-related work, testing, simulation, reduction to practice, and patent drafting. Patent rights still require a natural-person inventor. Routine AI assistance creates no automatic duty to announce AI use to the USPTO or EPO.

The risk arises when AI involvement affects inventorship, possession of the claimed invention, the accuracy of patent-office submissions, or the clarity of the inventor designation. The practical response for in-house counsel is process-based: ask about AI use early, document the human contribution, preserve important AI interactions, audit AI-generated drafting, and coordinate U.S. and European filing language before any optional AI-attribution statement is made.