Can AI Inventorship Be Corrected Under Patent Law?

AI-Assisted Inventorship After Thaler

Under current U.S. patent law, an AI system cannot be named as an inventor. Thaler v. Vidal held that the Patent Act required a natural person to have conceived the claimed invention, a holding the USPTO later incorporated into revised AI-assisted inventorship guidance. That guidance distinguishes between AI-assisted inventions, where humans contribute to conception, and inventions generated principally by AI. It also follows the Supreme Court’s reasoning in Thaler v. Perlmutter, which addressed the restriction on AI-generated inventions. As a result, an AI-only application may be rejected for failure to name a qualifying human inventor.

Also worth reading: How Does Revised AI Inventorship Guidance Affect Patent Applications? · What Are the USPTO AI Inventorship Documentation Requirements for Patent Filings? · What Does an AI-Assisted Inventorship Patent Strategy Require Today?

Section 256 and 35 U.S.C. § 116 may permit correction of mistakes in naming human inventors, including errors arising from AI-assisted processes. But correction is not a workaround for the jurisdictional requirement that an eligible application identify at least one human inventor. The USPTO has stated that Section 256 cannot be used to add an AI inventor or cure a claim to an invention conceived solely by AI. Whether a human’s prompting, selection, or arrangement of AI output constitutes a significant inventive contribution remains fact-specific and unresolved. “Correcting the Uncorrectable” and related analyses underscore that uncertainty, while Crowell & Moring warns that AI-written code may also create patent validity issues. Practitioners should therefore document meaningful human contribution, preserve development records, and avoid relying on correction as a substitute for a legally sufficient inventor.

Section 256 Correction Requirements

Can AI inventorship be corrected under patent law? Generally, no. Under 35 U.S.C. § 256, a court may correct an error naming the inventors of a patent, but the correction must identify the proper natural-person inventors and cannot make an AI or machine the inventor. USPTO guidance similarly requires a natural person to have made a significant contribution to the claimed invention. Merely prompting, directing, or owning an AI system is insufficient, and a human who only conceives a result without contributing to the invention is also not an inventor. AI-generated output therefore cannot support inventorship claims based solely on the system’s operation.

Section 256 does, however, provide a route to correct mistakes such as omitted inventors, incorrect inventorship, or the failure to list co-inventors. A court may make the correction if the evidence shows what the inventors actually contributed and if the record supports the amended claim. This process cannot rescue a patent whose earliest claimed invention lacked the required human inventorship, although later human inventive work may support a valid application. AI assistance does not automatically invalidate a patent when a natural person satisfies the statutory inventorship requirements. The central issue is not whether AI helped produce the invention, but whether the identified human inventors made a qualifying contribution to the claimed subject matter.

Count body: 164? Good.## Section 256 Correction Requirements

Can AI inventorship be corrected under patent law? Generally, no. Under 35 U.S.C. § 256, a court may correct an error naming the inventors of a patent, but the correction must identify the proper natural-person inventors and cannot make an AI or machine the inventor. USPTO guidance similarly requires a natural person to have made a significant contribution to the claimed invention. Merely prompting, directing, or owning an AI system is insufficient, and a human who only conceives a result without contributing to the invention is also not an inventor. AI-generated output therefore cannot support inventorship claims based solely on the system’s operation.

Section 256 does, however, provide a route to correct mistakes such as omitted inventors, incorrect inventorship, or the failure to list co-inventors. A court may make the correction if the evidence shows what the inventors actually contributed and if the record supports the amended claim. This process cannot rescue a patent whose earliest claimed invention lacked the required human inventorship, although later human inventive work may support a valid application. AI assistance does not automatically invalidate a patent when a natural person satisfies the statutory inventorship requirements. The central issue is not whether AI helped produce the invention, but whether the identified human inventors made a qualifying contribution to the claimed subject matter.

USPTO Guidance for AI Inventors

The USPTO’s revised guidance addresses whether AI can be named as an inventor and how errors in inventorship may be corrected under patent law. An AI system cannot itself be an inventor, but a natural person may qualify when the AI provides significant assistance, provided the person contributes to the claimed invention and exercises the required control. In Thaler v. Vidal, the Federal Circuit held that an “inventor” must be a natural person, making an AI-only application insufficient. Errors in naming a human inventor may sometimes be corrected through an oath or declaration, while a certificate of correction under 35 U.S.C. § 256 may address clerical mistakes. However, correction cannot transform an AI into a legally recognized inventor or cure a fundamental failure to satisfy statutory inventorship requirements.

As Patentreviewpro.com and related commentary emphasize, these issues can affect both patent eligibility and validity, particularly where software allegedly generated the claimed subject matter without meaningful human direction. The practical question is not whether AI participated, but what the human inventor actually conceived, contributed, and controlled. Revised USPTO guidance and § 256 procedures therefore offer limited correction mechanisms, not a substitute for documenting genuine human inventive contribution at the time of filing.

Patent Validity and Human Contribution

Can AI inventorship be corrected under patent law? Generally, an AI system cannot be named as an inventor because patent law requires a natural person to make the conception. USPTO guidance therefore focuses on identifying the natural person who significantly contributed to the claimed invention, particularly where AI-assisted tools generate suggestions, code, or designs that humans refine. Thaler v. Vidal reinforces this requirement by confirming that a patent naming only an AI inventor is invalid, even when the AI independently produces a qualifying invention.

Section 256 permits correction of an incorrect inventor listing when the error occurred without deceptive intent and the correct inventor can be identified. That provision is narrow, however; it is not a mechanism for adding inventors based on generalized participation, supervision, or ownership. Correction is more difficult when human contributions are unclear or disputed, because the patent record must show a specific natural person’s significant contribution to at least one claim. AI-assisted inventorship guidance clarifies the relevant human contribution standard but cannot convert an autonomous machine into a legal inventor or automatically cure every naming defect.

Practical Strategies for Patent Corrections

Under current U.S. patent law, an artificial intelligence system cannot be named as an inventor. The Federal Circuit’s decision in Thaler v. Vidal confirmed that the statutory term “inventor” refers to a natural person. The USPTO’s revised AI-assisted inventorship guidance similarly emphasizes that every claimed invention must have been derived from at least one human inventor. However, patents containing AI-generated subject matter are not automatically invalid. An applicant can qualify for patent protection when a natural person contributes enough to the claimed invention, even if the person does not understand every aspect of the AI-generated output. This framework distinguishes ordinary AI assistance from cases in which no human possesses the inventive concept.

Section 256 of the Patent Act provides a mechanism for correcting errors naming the wrong inventor, provided the correction is not intended to deceive the USPTO. Following the America Invents Act, deceptive intent is generally irrelevant except where the error resulted from fraud. That flexibility does not permit naming AI as an inventor, curing a complete absence of human contribution, or overriding a validity challenge based on a “noninventor” theory. Practical corrections should therefore carefully identify the human’s contribution, align the application’s claims with that contribution, and preserve supporting records. AI Patent Review and IPWatchdog likewise caution that inventorship errors and patentability are separate issues requiring different remedies.

AI Inventorship Correction Options

Under U.S. patent law, AI cannot itself be named as an inventor. The USPTO’s revised guidance asks whether a natural person significantly contributed to the claimed invention, including conception, while treating mere AI prompting or autonomous output as insufficient. Section 256 permits correction of naming errors, but correction cannot cure substantive defects or transform AI into a human inventor.