By: Danielle Kimel
Published: August 9, 2025
Despite popular depictions of Artificial Intelligence[1] (AI), systems that can think, feel, and learn like a human are still fictional.[2] Nonetheless, does a creator have to be a human being for their work to receive intellectual property protection? For patent and copyright, courts agree that the answer is no.[3] Although there is no relevant case law, the language of trade secret statutes implies that AI-generated information can receive protection as trade secrets.[4]
When addressing AI and intellectual property, copyright infringement is often the first topic that comes to mind, due to many AI systems being trained on copyrighted materials.[5] However, providing intellectual property protection for AI outputs is just as important as preventing the use of AI as a tool for infringement. The question we are presented with is what kind of intellectual property protection, if any, can apply to AI output?
Whether an AI system can be the author of a copyrightable work was only clarified recently, in Thaler v. Perlmutter.[6] Perlmutter was a test case filed by AI advocate Stephen Thaler to see how U.S. courts would address purported AI-artists.[7] The case turned on the Copyright Act of 1976’s requirements for a valid copyright claim, which are that the work be an original work of authorship fixed in a tangible medium of expression.[8] In plain English, this means that a copyrightable work must be original, it must be the work of an author, and it must be recorded in a tangible way such that it can communicated to others.[9] Perlmutter’s holding turned on whether an AI, operating without a guiding human hand, constitutes an author for the Copyright Act’s purposes.[10] The Perlmutter court concluded that human authorship is a “bedrock requirement of copyright.” Therefore, an AI cannot author a copyrightable work, and the work in question was correctly denied copyright protection.[11]
Similarly, the 2022 case Thaler v. Vidal held that a valid patent claim must have a human inventor; therefore, any patent claim submitted to the United States Patent and Trademark Office (USPTO) with an AI as the named inventor should be denied.[12] Vidal was another test case filed to see how the courts address purported AI-inventors. In this case, the Federal Circuit unambiguously stated that the Patent Act requires that inventors be human beings.[13] Therefore, an AI cannot be the named inventor of a patentable invention.[14]
Copyright and patent law both inherently require a human creator, making them incompatible with protecting AI output.[15] This is where trade secrets come in, as a form of intellectual property uniquely suited to protecting valuable information generated by AI systems without human guidance.[16]
Although there have been no cases so far testing the eligibility of AI outputs for trade secret protection, trade secrets and AI are likely to be a winning combination. Unlike copyright and patent law, which require a human creator, the key requirement of a valid trade secret claim is secrecy, regardless of who or what created the trade secret.[17] Statutory language strongly implies that (1) whether a human being created the trade secret is irrelevant, and (2) AI-generated information can qualify as a protectable trade secret if the information has inherent economic value and the owner of the secret has taken reasonable measures to keep it secret.[18]
In plainer language, if a person has a legal right to valuable information that is not generally known or readily ascertainable and takes reasonable steps to keep that information secret, then that information can be a trade secret. Whether or not a human being created the trade secret never comes into the equation; even the term “owner” in trade secret law can refer to either a person or an entity.[19] Thus, trade secrets are likely the best protection available for AI-outputs, so long as the output meets the value and secrecy requirements.
For example, a hypothetical pharmaceutical company has several cancer therapies. The company has an AI trained on the company’s database, and they ask their AI to develop a new treatment. The AI produces a new, more effective therapy—saving many lives—and the company immediately takes steps to keep the information confidential. This new therapy could constitute a valid and protectable trade secret because it is (1) information that (2) has inherent economic value that comes from the information being kept secret from the general public and competitors, and (3) the owner has made reasonable efforts for their circumstances to keep the information secret.[20] Nowhere in the elements required for a trade secret does the creator of the trade secret come into play. In this situation, it would not matter if the trade secret was developed by a non-human entity, it only matters that the information is valuable because of its secrecy and is kept secret.
Protecting trade secrets developed by AI is important for fostering innovation and economic equity. AI systems can easily generate information not originally in the possession of human beings, which has economic value that is maintained by its secrecy. If, for example, an individual or small business owns valuable AI-generated information but cannot claim the information as a trade secret, then there would be no reason for their competitors not to take and use that valuable information for their own gain. Individuals and small businesses would struggle to compete with larger competitors, and they would have no legal recourse, since they would likely be unable to protect that same information through copyright or patent protection.[21]
Extending trade secret law to protect AI-generated information levels the playing field, ensuring that those who invest in AI can benefit from the AI-generated trade secrets without fear of unfair competition.[22] Moreover, if AI-generated trade secrets are not protected, then there will be less motivation to use AI as a tool for innovation, despite its many benefits.[23]
[1] In this article, Artificial Intelligence, or AI, refers to deep-learning models designed to generate high-quality content based on the data the system is trained on. This type of AI is also called Generative AI.
[2] See generally 2001: A Space Odyssey (Stanley Kubrick dir., 1968) (featuring an AI, Hal 9000, that turns on the main character).
[3] See Thaler v. Perlmutter, 687 F. Supp. 3d 140, 150 (D.D.C. 2023) (holding that AI cannot be the author of a copyrightable work; without a human author a work cannot be copyrighted); Thaler v. Vidal, 43 F.4th 1207, 1212 (Fed. Cir. 2022) (holding that AI cannot be the named inventor of a patentable invention; patent applications with AI as the inventor will be denied for improper inventorship).
[4] See Trade Secret, Black’s L. Dictionary (12th ed. 2024) (defining “Trade Secret” as “[a] formula, process, device, or other business information that is kept confidential to maintain an advantage over competitors”); Defend Trade Secrets Act of 2016, Pub. L. 114-153 (May 11, 2016) (defining trade secrets and misappropriation, and setting out a private cause of action for trade secret misappropriation); 18 U.S.C. § 1839 (implying with its definition of “trade secret” that the validity of a trade secret turns on the value and secrecy of the information, not the source of the information).
[5] See, e.g., Thomson Reuters Enterprise Centre GMBH v. Ross Intelligence, Inc., No. 1:20-CV-613-SB, 2025 WL 458520, at *10 (D. Del. Feb. 11, 2025) (ruling that an AI developer’s use of copyrighted material to train the AI system was not fair use as a matter of law).
[6] Perlmutter, 687 F. Supp. 3d at 150 (holding that because human authorship is an essential part of a valid copyright claim, the AI-generated work in question did not rise to a valid copyright upon its creation).
[7] Id. (suing the U.S. Copyright Office for its denial of Thaler’s copyright application, which listed Thaler’s AI, the Creativity Machine, as the author); The Artificial Inventor Proj., Copyright, https://artificialinventor.com/copyright/ (last visited Mar. 10, 2025) (listing Thaler v. Perlmutter as one of the Artificial Inventor Project’s pro bono legal test cases seeking intellectual property rights for AI-generated output).
[8] 17 U.S.C. § 102.
[9] Id.
[10] Perlmutter, 687 F. Supp. 3d at 146.
[11] Id. at 146-47 (noting that the presumption of the authorship requirement being a requirement of a human author rests on centuries of settled understanding that the act of human creation lies at the center of American copyright law; “non-human actors need no incentivization with the promise of exclusive rights. . .and copyright law was therefore not designed to reach them.”); see also Naruto v. Slater, 888 F.3d 418, 420 (9th Cir. 2018) (holding that the photo a monkey had taken of itself was not copyrightable, because “all animals, since they are not human,” lacked statutory standing under the Copyright Act).
[12] Thaler v. Vidal, 43 F.4th 1207, 1213 (Fed. Cir. 2022) (holding that only a “natural person,” meaning a human being, can be an inventor, so AI cannot be an inventor; affirming the lower court’s decision to reject Thaler’s patent claim listing his AI as the sole inventor).
[13] Id. at 1211 (citing Mohammad v. Palestinian Auth., 566 U.S. 449, 454 (2012)) (stating that the Patent Act expressly provides that inventors are “individuals” and that the Supreme Court defined “individual,” when used as a noun, to ordinarily mean a human being).
[14] See id at 1213 (“here, Congress has determined that only a natural person can be an inventor, so AI cannot be”).
[15] See id.; Perlmutter, 687 F. Supp. 3d at 150.
[16] This article understands “trade secret(s)” to mean information, including a formula, pattern, compilation, program, device, method, technique, or process that derives independent economic value, actual or potential, derived from its secrecy, and is the subject of reasonable efforts to maintain that secrecy. A common example of a trade secret is the recipe for Coca-Cola or KFC’s secret blend of herbs and spices.
[17] See 18 U.S.C. § 1839.
[18] See § 1839(4) (defining a trade secret “owner” as the person or entity holding rightful legal or equitable title to, or license in, the trade secret; implying that the owner of the trade secret is not necessarily the creator of the trade secret); § 1839(3) (defining a valid trade secret as (1) information that (2) has independent economic value, and (3) which the owner has taken reasonable measures to keep secret).
[19] § 1839(4).
[20] § 1839(3) (defining a valid trade secret as (1) information that (2) has inherent economic value, and (3) which the owner has taken reasonable measures to keep secret).
[21] See generally Thaler v. Perlmutter, 687 F. Supp. 3d 140 (D.D.C. 2023) (holding that AI cannot be the author of a copyrightable work; without a human author a work cannot be copyrighted); Thaler v. Vidal, 43 F.4th 1207 (Fed. Cir. 2022) (holding that AI cannot be the named inventor of a patentable invention; patent applications with AI as the inventor will be denied for improper inventorship).
[22] See generally Jessica Maliya Myers, Artificial Intelligence and Trade Secrets, A.B.A. (Jan. 16, 2019), https://www.americanbar.org/groups/intellectual_property_law/resources/landslide/archive/artificial-intelligence-trade-secrets/ (stating that benefits of trade secret protection include immediate protection without formal registration or novelty requirements, and have always been important to protecting competitive business information; noting that Congress’s enacting the DTSA was influenced by reports of trade secret theft reaching unprecedented levels); Ekene Chuks-Okeke, et al., Generative AI and intellectual property: The evolving copyright landscape, IAPP (July 31, 2024), https://iapp.org/news/a/generative-ai-and-intellectual-property-the-evolving-copyright-landscape (arguing that intellectual property rights protect business assets, but also provide protection for individual creators not associated with a business enterprise).
[23] See generally Myers, Artificial Intelligence and Trade Secrets, supra note 22 (remarking that “AI requires significant investments,” implying that failure to protect AI output will discourage use of AI as a creative tool because the cost will outweigh potential benefit; also noting that in countries without strong legal protection or enforcement of trade secrets, companies place restrictions on the flow of information between potential business partners that slow down innovation, creating concerns that even the companies owning the trade secrets suffer from reduced innovation); Joseph Barber, Trade Secrets and AI Systems: The Future of Trade Secret Protection in Business, IPWatchdog (Dec. 12, 2024, 1:15 PM), https://ipwatchdog.com/2024/12/12/trade-secrets-and-ai-systems-the-future-of-trade-secret-protection-in-business/id=184002/ (commenting that “developing AI tools and systems that are adopted by a critical mass in general or in a specific niche can result in an astronomical value to the developing company”).