We are in the age of artificial intelligence. AI tools have quickly become part of everyday work, including research and development. In fact, I used an AI tool to assist with the research for this article. But don’t worry, I double-checked all the sources and confirmed the accuracy of the AI-generated output. While AI tools are great research assistants and helpful in brainstorming ideas, users should be cautious when using AI to avoid inadvertently jeopardizing their intellectual property rights. In this article, I discuss some of the implications of using AI tools for patent and trade secret protection.
For both patent protection and trade secret protection, it is important to keep certain information confidential. For patents, an invention generally cannot be patented if it was “available to the public” before the patent application was filed (known as a “public disclosure”). See 35 U.S.C. § 102. Courts have held that sharing information with others who have a reasonable expectation of confidentiality does not make the information publicly available, even without a formal confidentiality agreement. See, e.g., Cordis Corp. v. Boston Scientific Corp., 561 F.3d 1319, 1333–35 (Fed. Cir. 2009); Dey, L.P. v. Sunovion Pharms., Inc., 715 F.3d 1351 (Fed. Cir. 2013). While the U.S. provides a one-year grace period for certain disclosures made by the inventor, most foreign jurisdictions require “absolute novelty” and will refuse patent protection for an invention that was publicly disclosed before the application was filed. For trade secrets, information qualifies as a “trade secret” only if, among other requirements, “reasonable measures” are taken “to keep such information secret.” See 18 U.S.C. § 1839(3)(A).
Inputting information into AI tools raises a question about whether the information has become “available to the public” or is no longer a secret. For patents, this question hinges on whether there is a reasonable expectation of confidentiality when sharing information with an AI tool. For trade secrets, the question is whether sharing the information with an AI tool is consistent with taking reasonable measures to maintain secrecy. When information is input into an AI tool, that information is sent to a third party’s servers. Depending on the provider’s terms, it may be stored, reviewed by the provider’s personnel, or used to train future versions of the model.
Since we are only at the early stages of the AI era, the law is not yet settled as to whether use of AI tools constitutes a public disclosure of the information submitted to the platforms, thereby forfeiting potential patent and trade secret rights.
However, certain cases are informative as to whether information input into an AI tool remains confidential and suggest that courts may treat the input of information into an AI tool, without confidentiality protections or no-training terms, as a public disclosure. While not binding precedent, at least one federal court has held that sharing information with a public AI tool, without confidentiality protections, can extinguish trade secret rights. See Trinidad v. OpenAI, Inc., No. 25-cv-06328-JST (N.D. Cal. Jan. 5, 2026) (dismissing a pro se plaintiff’s trade secret claim because, by developing her ideas through ChatGPT, the plaintiff voluntarily disclosed them to the defendant, the developer of ChatGPT).
In another recent case, a district court found that communications made through a generative AI platform were not confidential and refused to apply the attorney-client privilege, which protects certain confidential communications from disclosure. See United States v. Heppner, 820 F. Supp. 3d 292, 296 (S.D.N.Y. 2026). This finding was based not merely on the fact that the communications were with a third-party AI platform, but on the provider’s privacy policy, which allowed it to collect inputs, train on them, and disclose them to third parties. See id. Based on the defendant’s consent to the platform’s policy, the court held that the defendant had “no reasonable expectation of confidentiality in his communications” with the AI platform. See id. at 297.
Although the above cases address trade secrets and privilege, the same reasoning could be applied to questions of public disclosure in patent cases. Where there is no reasonable expectation of confidentiality in communications with an AI tool, a court could find that the information was made available to the public, constituting a public disclosure. Until courts resolve this question, the safest course is to assume that such communications may be treated as a public disclosure.
Based on the current legal landscape, it is best practice to avoid entering sensitive information, such as patentable subject matter or trade secrets, into free consumer versions of AI tools, which typically reserve the right to use inputs for training, allow the provider’s employees to review inputs, or permit the provider to disclose inputs to third parties. With such tools, there may be no reasonable expectation of confidentiality, and sharing information could result in a loss of patent rights, trade secret protection, or both. While the law remains unsettled, use of an AI tool with strong contractual confidentiality protections (often available in certain enterprise versions) and with no training on your inputs appears to pose less of a risk of a public disclosure and loss of IP rights.
