AI can't hold patents to U.S. inventions (for now)

Semiconductor chips are seen on a circuit board of a computer in this illustration picture taken February 25, 2022. REUTERS/Florence Lo/Illustration

October 20, 2022 - Three years ago, Stephen Thaler filed two patent applications naming a single inventor, an Artificial Intelligence (AI) program. The U.S. Patent and Trademark Office (USPTO), following Director review, found the applications to be incomplete for lacking a valid inventor on the ground that a machine cannot be an inventor. Thaler appealed the USPTO's final decisions to the District Court, which similarly concluded that an "inventor" must be a natural person.

Thaler then appealed to the U.S. Court of Appeals for the Federal Circuit, with the sole issue being whether AI qualifies as an "inventor" under U.S. patent law. The answer (for now) is no. (Thaler v. Vidal, No. 21-2347, *2 (Fed. Cir. Aug. 5, 2022)).

Federal Circuit’s rationale

The Court first looked to the definition of "inventor" in the Patent Act, which has defined the term as an "individual" since 2011. (Id. at *6). While the Patent Act does not further define "individual," the Court was guided by both the ordinary meaning of the term and related case law. (Id. at *6-7). The Court cited to the Supreme Court's explanation that the noun "individual" ordinarily means a human being. (Id. at *6).

This explanation was also consistent with how the term is used in "everyday parlance" and dictionary definitions of the term. (Id). The Court also referred to Supreme Court precedent holding that the word "individual" in statutes refers to human beings unless there is some evidence that Congress intended otherwise. (Id. at *7). Looking to the rest of the Patent Act, the Court found no such evidence. (Id).

While Thaler relied on other provisions of the Patent Act, the Court found no need to turn to these provisions when the Act included an explicit provision addressing who may be an inventor — an individual. (Id. at *7-9). Responding to Thaler's insistence that the Court should consider the "broader context of the statute as a whole," the Court found that the Patent Act as a whole points to an "inventor" being a human being. (Id. at *9).

Mixed holdings on permissibility of AI as an inventor internationally

The U.S. is not the only country that does not allow AI to be an inventor. A general appellate court in the United Kingdom also recently decided against AI inventorship for Thaler's patents. The U.K. Court of Appeal (Civil Division) similarly looked to the country's patent statute, under which an "inventor" must be a "person" to be an actual deviser as required in the statute.

However, in Australia, the Full Court of the Federal Court, which is a general court whose appeals can only be heard at the High Court, found that AI could be an "inventor" because Australian law did not preclude such an interpretation, and the court recognized "the evolving nature of patentable inventions and their creators." The South African Patent Office, which has no substantive patent examination system, has granted Thaler's patent naming AI as the inventor.

Implications of this decision

So what happens next? For now, Thaler has requested a rehearing of the case at the Federal Circuit. (Thaler v. Vidal, No. 21-2347, Paper No. 62 (Fed. Cir. Aug. 5, 2022)). In his request, Thaler argues that (1) the dictionary definition of "individual" encompasses AI; (2) denying patent protection for AI's inventions contravenes Supreme Court precedent against denying patents based on how inventions are made; and (3) the decision contradicted the Supreme Court's statutory interpretation approach in view of evolving technology. (Id. at 10-11).

The Federal Circuit's nominal bright line rule requiring an "inventor" to be a human provides clarity on the front end, supporting the USPTO's decision to reject patent applications where the inventor is declared to be AI. The record on appeal allowed the Court to resolve a pure question of law. That is because the applicant was declared to be AI. The Court never reached the more complex question of what level of reliance on AI in the invention process would deprive a human from claiming inventorship.

Put another way, if AI is used as a tool to invent, then, like traditional inventions, the use of tools to carry out an invention does not bar an applicant's claim of inventorship. But, under the current Federal Circuit ruling, if the role of AI in the inventive process rises to the level of an inventor, then the invention is ineligible for patent protection. This is a factual question with perhaps some very difficult line drawing that was avoided by the Thaler case.

Note that this type of line drawing is not unprecedented. Our analysis of inventorship under U.S. law routinely requires consideration of whether the contributions of human collaborators make them an inventor. If so, the humans are named on the patent application as inventors and prosecution of the patent moves forward. Under Thaler, however, if the same inventorship analysis concludes that AI is an inventor, the AI claims cannot be patented.

Design patent drafters may face similar issues, because the same inventorship statute applies to design and utility patent applications. As with utility inventions, designers have regularly used computer programs (such as CAD) as tools to aid in reducing their designs to practice. But increasingly AI is being used to optimize existing designs (for a variety of reasons, but at least one is to reduce the weight of material needed to make a product).

When these changes affect the appearance of the design, the AI may be seen as taking on a co-designer role. We may soon reach a time when the design is completely created by AI, with humans making only the final design selection.

Similar issues exist in copyright law. Earlier this year, the Copyright Office again found that a copyright could not be issued for art generated by AI, because such a work lacked "human authorship." (Notably, this case also involved Thaler.) While Thaler argued that this holding was unconstitutional, the Copyright Office refused to "depart from a century of copyright jurisprudence" on this point, including a decision from 2018 concluding that a monkey could not sue for copyright infringement for a photograph it took after stealing a photographer's camera.

Possible ways forward

To address this potential AI patentability bar, there are several strategies available to innovators to protect IP. When drafting a utility patent specification directed to AI-related technology, the disclosure should clearly describe the AI technology as a tool used by humans to implement the invention. The claims in patent applications, the touchstone for the inventorship analysis, should be drafted such that each of the elements are directed to human contributions.

If it is not possible or desirable to disclose and claim the invention in a way that relegates AI to a tool and focuses on human invention, other forms of IP protection are available. One option is trade secret protection. The obligation barring public disclosure to maintain trade secret status means that inventors do not have to comply with the disclosure and claiming obligations of the patent statute. This can have particular benefits when it comes to the challenges of meeting those obligations posed by AI inventions.

Furthermore, there is no inventorship determination required to protect trade secrets. So, it is irrelevant for purposes of determining whether or not an invention qualifies for trade secret protection whether it originates from human or non-human contributions. Accordingly, because the statutory framework particular to patent protection that resulted in the Thaler test does not exist, the Thaler bar does not apply to trade secret protection.

For designs, it may still be possible to register for trade dress rights in the design, provided that the design meets the non-functionality requirement and has developed secondary meaning. Trade dress is a type of trademark that does not require identification of a human designer.

While we wait for further guidance from the Supreme Court, this case may act as motivation for Congress to consider the Patent Act (and perhaps the Copyright Act) in light of evolving technology. This task is not without its challenges. Congress would also have to modify the requirement to declare that an inventor believes they are the original inventor.

Also, the written description requirement may change if it is geared towards a disclosure sufficient for a machine to understand instead of (or in addition to) a human. Still, allowing AI inventorship may help avoid a bias towards trade secret protection for AI that could unnecessarily reduce the knowledge placed in the public domain. And if Congress later changed the statute, there could be a "bubble" of AI-created inventions that are no longer eligible for patenting (due to, for example, the on-sale bar).

But cutting against all these considerations is the question of whether AI even needs the protection of inventorship. AI inventions may be relatively cheap to generate, cutting against the incentive to invest that might be driven by the availability of patent rights. But these inventions could potentially be quite valuable and thus vulnerable to infringement by competitors.

Finally, the question of property rights for machines adds an additional layer of uncertainty to these rights. So it seems that Thaler may be the first case in a long road considering the many aspects of this multifaceted issue.

Opinions expressed are those of the author. They do not reflect the views of Reuters News, which, under the Trust Principles, is committed to integrity, independence, and freedom from bias. Westlaw Today is owned by Thomson Reuters and operates independently of Reuters News.

Angeline Premraj is an associate in the Washington, D.C., office of Finnegan, Henderson, Farabow, Garrett & Dunner LLP. Her diverse patent practice includes client counseling, prosecution, litigation, and post-grant proceedings before the Patent Trial and Appeal Board (PTAB) of the U.S. Patent and Trademark Office (USPTO). Her technical background is rooted in medical devices, and she also works with a broad field of technologies, including augmented and virtual reality (AR/VR), electrical and computer technology, financial and business systems, and manufacturing. She can be reached at

Elizabeth Ferrill, a partner in the Washington, D.C., office of Finnegan, Henderson, Farabow, Garrett & Dunner LLP, focuses her practice on all aspects of design patents, including prosecution, counseling, post-grant, and litigation. She can be reached at

Frank A. DeCosta III is co-leader of Finnegan, Henderson, Farabow, Garrett & Dunner LLP’s AI/ML section, manages patent enforcement activities for clients in the United States, Europe, and Asia and is involved in all phases of patent litigation. He provides opinions related to computers, consumer electronics, software, medical devices, and information systems and is lead counsel in appeals before the U.S. Court of Appeals for the Federal Circuit and in inter partes review proceedings before the Patent Trial and Appeal Board. His research, academic training, and work experience involve hardware and software research and development of multimedia telecommunications systems, complex signal processing systems, image processing systems, medical systems, and pattern recognition systems. He has a doctorate in electrical engineering, with a focus on neural network-based AI systems for image recognition. He can be reached at