The unique analytical power and output of AI tools have raised entirely new questions around patent inventorship. Future patent litigation is likely to see alleged infringers asserting a defense of patent invalidity on the grounds that there is no human inventor. This new risk consideration may also begin impacting diligence and valuation assessments. Patent owners, applicants, and parties involved in evaluating patent portfolios should account for this risk and evaluate potential proactive protections.
Two underlying concepts are key for navigating inventorship in the AI era:
- An AI cannot be an inventor. Thaler v. Vidal, 43 F.4th 1207, 1211-1212 (Fed. Cir. 2022). The court concluded that “Congress has determined that only a natural person can be an inventor, so AI cannot be.” Id. at 1213. It is worth noting that the court expressly left open “whether inventions made by human beings with the assistance of AI are eligible for patent protection.” Id.
- A patent that does not name the correct inventors is invalid unless the error can be corrected under 35 USC §256. Fortress Iron, LP v. Digger Specialties, Inc., No. 24-2313, slip op. at 11 (Fed. Cir. Apr. 2, 2026); Pannu v. Iolab Corp., 155 F.3d 1344, 1349-50 (Fed. Cir. 1998).
This means that, for a claim to be patentable in the US, it must have at least one human inventor.
It is well established that an inventor is a human individual who conceives the invention recited in at least one claim of a patent, and conception can be performed either individually or cooperatively, as discussed in the Manual of Patent Examining Procedure (MPEP) §2109. Conception is the process of forming “a definite and permanent idea of the complete and operative invention, as it is hereafter to be applied in practice,” according to Hybritech Inc. v. Monoclonal Antibodies, Inc., 802 F.2d 1367, 1376 (Fed. Cir. 1986) (quoting 1 Robinson on Patents 532 (1890)).
Conversely, it is also well established that it is not inventorship to:
- Suggest “an idea of a result to be accomplished, rather than the means of accomplishing it,” (Ex parte Smernoff, 215 USPQ 545, 547 (Bd. App. 1982)), as quoted in MPEP §2109;
- “[M]erely explain to the real inventors well-known concepts and/or the current state of the art” (Pannu v. Iolab Corp., 155 F.3d 1344, 1351 (Fed. Cir. 1998)) or provide a “contribution of information in the prior art” (Eli Lilly and Co. v. Aradigm Corp., 376 F.3d 1352, 1362 (Fed. Cir. 2004)); or
- Only carry out the plans or directions of an inventor or only carry out steps that require no more than ordinary skill, as discussed in MPEP § 2109(III); Fritsch v. Lin, 1991 WL 332571, at 3 (Bd. Pat. App. & Inter. 1991); see also Fina Oil & Chem. Co. v. Ewen, 123 F.3d 1466, 1473 (Fed. Cir. 1997).
In November 2025, the US Patent and Trademark Office (USPTO) issued revised inventorship guidance for AI-assisted inventions, making clear that the same legal standard applies to all inventions regardless of whether AI was used. Under that guidance, AI is treated as a tool used by the human inventor, much like laboratory equipment or software. Courts have not yet addressed this approach, and USPTO guidance does not bind them.
The risk
As inventors increasingly rely on AI to research, analyze data, and answer questions, there is an increasing risk that (a) no human will conduct conception and/or (b) records of AI use will give accused infringers evidence to challenge human conception. Issued patents are presumed to name the correct inventors, and a challenger must prove incorrect inventorship by clear and convincing evidence. But if the record shows that AI, rather than a human, conceived the claimed subject matter, one or more claims may be found invalid.
Patent owners should also be aware that there is a risk that a single claim lacking human inventorship could invalidate the entire patent, including human-invented claims. This question has not yet been directly addressed in the courts, but the Federal Circuit’s treatment of uncorrectable inventorship in Fortress Iron raises the possibility that claims lacking human inventorship could function as “poison pills” for an entire patent, in an exception to the usual claim-by-claim approach to validity (35 USC §282(a); 35 USC §253(a)).
The goal
A forward-looking patent owner should ensure that every patent claim has at least one human inventor and document that inventor’s contributions to withstand future litigation.
The mitigation
Patent owners can adapt their initial invention review and triage procedures to establish and document human inventorship when AI tools were used. Creating this record while the relevant individuals are available, their recollections are fresh, and supporting records are accessible can help defend against later legal challenges. Patent owners should consider incorporating some or all of the following into their initial disclosure processes:
- Modify disclosure forms to ask whether AI was used in the project.
- If AI was used, the inventor(s) should be asked to provide evidence of at least one of the following:
- Evidence of a prompt that was specific enough that the AI is doing no more than carrying out a human’s plans or directions.
- For example, if researchers give an AI a specific selection of cancer patient gene expression data sets along with a prompt which includes specific instructions to find genes with a certain fold-change, a certain degree of correlation with some pathological metric, or which meet a certain multi-factorial threshold with specific weights for each factor, it can be argued that the AI is acting as no more than a tool carrying out the researchers’ plan when it conducts the analysis and provides a ranked list of potential targets.
- Evidence that one or more humans selected, narrowed, or modified AI-generated output to conceive an invention that goes beyond those results.
- For example, if researchers provide a broad, unspecific prompt to an AI, such as “identify likely therapeutics based on all cancer patient gene expression data sets you can find,” a patent owner should be looking for inventive human activity post-AI, such as selecting leads from the AI results based on additional knowledge or testing.
- Evidence of a prompt that was specific enough that the AI is doing no more than carrying out a human’s plans or directions.
The value of such policies should be balanced against the risk that the records will be discoverable and may be mined by accused infringers for evidence that the AI did the conceiving. As always, best practices dictate that records are accurate and contemporaneous, prepared with counsel involved, where possible, and aligned with retention policies.
AI-era triage: A new addition to the category of at-risk inventions
Savvy patent owners must also confront the reality that some inventions will not have a human inventor and, therefore, cannot be validly patented. Much like acknowledging scenarios in which a public disclosure has forfeited patent rights, recognizing AI-only inventorship at the outset can protect the patent owner’s ultimate financial and strategic interests. In such cases, patent owners should carefully review the technology and consider whether trade secret protection is a viable alternative or whether further human improvements or modifications could support a patent claim with human inventorship.
AI risk is not created equal: A new consideration for public disclosures
When asking inventors about their AI usage, patent owners may also want to ask which AI technology was used, and under what terms. Use of some AI technologies, particularly non-proprietary, publicly accessible AI tools, including those whose terms allow the provider to retain or train on user inputs, may undermine trade secret protection and create discoverable records. Whether that use could also count as a public disclosure for patent purposes has not been decided. In the US, the question is whether the information became accessible to the public. The risk is greater in jurisdictions that require absolute novelty.
Thus, during an initial triage analysis of an invention, patent owners should consider whether researchers’ AI use has put confidentiality or patent rights at risk.
Takeaways
- Establish policies and training to limit the use of non-proprietary, publicly accessible AI tools and to structure AI use so that the human contribution is substantive and documented.
- Conduct human inventorship clearance upon internal disclosure of a potential invention.
- Document human inventorship where AI has been used.
- Consider alternative IP protection strategies when human inventorship is not clearly established.
- Draft claims around the human contribution, confirm inventorship claim by claim, and cancel any claim lacking a human conceiver before issuance.
