Who Invented That? When AI Writes the Code, Patent Validity Issues May Follow

In Fortress Iron, LP v. Digger Specialties, Inc., No. 24-2313 (Fed. Cir. Apr. 2, 2026), the U.S. Court of Appeals for the Federal Circuit reaffirmed what happens when a patent incorrectly lists the true inventors, and that error cannot be corrected under 35 U.S.C. § 256(b), which requires notice and a hearing for all “parties concerned.” In Fortress, the patent owner sought judicial correction to add an inventor under § 256(b), but that inventor could not be located. Because the missing inventor qualified as a “concerned” party under the statute, the lack of notice and a hearing for that inventor made correction under § 256(b) impossible, and the patents could not be saved from invalidity.

The court grounded its invalidity finding in the “necessary and opposite implication” of § 256(b): if the error cannot be corrected, the patent is invalid. The court confirmed that this rule survives the America Invents Act’s (AIA) repeal of former § 102(f), and rejected the argument that naming at least one inventor is sufficient. Reading §§ 100(f), 101, and 256(b) together, all coinventors must be named — and where correction is legally or practically impossible, the savings clause offers no rescue.

This decision may have significant implications for AI-assisted innovation. AI cannot be named as an inventor under U.S. law. See Thaler v. Vidal, 43 F.4th 1207, 1210 (Fed. Cir. 2022) (holding that “inventors must be human beings”). But AI is increasingly being used in the invention process to assist human inventors. While it may be possible to correct inventorship and add a human inventor under the statute, what happens when a patent mistakenly names one or more human inventors, but litigation discovery reveals that the claimed invention was conceived entirely by AI? Unlike in Fortress Iron, where correction may have been obtained had the missing inventor been found and given an opportunity to be heard, AI cannot be named as an “inventor” as a matter of law. In such cases, there would appear to be no permissible inventor to substitute and no way to save the patent from invalidity.

The Federal Circuit’s decision in Fortress Iron is a reminder regarding the importance of (i) properly assessing and identifying inventors, and (ii) the risk that an incorrect inventorship determination up front may potentially lead to an invalid patent. Patentees should document human inventive contributions carefully and contemporaneously during the invention process, and work with the inventors to ensure that they obtain patent claims that each reflect human invention (lest an AI contribution be deemed to render claims unpatentable). Further, companies with patent portfolios may consider auditing pre-existing patent portfolios where AI was used in the invention process, to identify and correct any potential issues. On the other hand, parties accused of patent infringement are advised to consider improper inventorship as a potential defense during litigation. A well-crafted discovery strategy concerning the use of AI during the invention process could potentially result in a finding of patent invalidity, particularly if it turns out the asserted claims are not the result of human invention.