Court dismisses trade secret claim based on information developed in a public AI tool
A federal court has dismissed a trade secret claim based on information the plaintiff developed while using a public artificial intelligence (AI) tool. In Trinidad v. OpenAI Inc., No. 25-cv-06328-JST (N.D. Cal. Jan. 5, 2026), Judge Jon S. Tigar of the Northern District of California dismissed with prejudice every claim brought by the pro se plaintiff, who alleged that OpenAI took AI “frameworks” she developed while using its public AI tool and incorporated them into its products, including as an agent and other features.
To prove a trade secrets case, a plaintiff must prove that the claimed secret was subject to at least reasonable efforts under the circumstances to preserve its secrecy. But Trinidad’s trade secret claim failed on the “reasonable measures” element under the Defend Trade Secrets Act. Building her ideas in the public AI tool “would have required her to voluntarily share” them with OpenAI, a party with no duty to keep them confidential. The court concluded that this disclosure was inconsistent with reasonable measures to maintain secrecy.
In reaching its decision, the court also addressed two arguments:
- Ownership is not secrecy: OpenAI’s Terms of Use (Terms) gave her ownership of her outputs, but the court held that ownership does not satisfy the secrecy requirement.
- Challenging the Terms did not undo the disclosure: The court found she consented “whether or not [her] consent is enforceable as a contractual matter.”
The court dismissed her remaining claims on other grounds. While Trinidad is a non-binding ruling, it applies established trade secret principles. Our alert provides key points and considerations for businesses.
AI use may become a focus in future trade secret cases
The court ruled on the pleadings before any discovery, based on the plaintiff’s own account of how she worked. Defendants may now argue, as a threshold matter, that an asserted secret passed through a publicly available AI tool whose terms allow provider use. Plaintiffs, in turn, may need to plead or prove facts showing their asserted secrets were not entered into such tools. A single employee’s use of a personal account could undermine a claim. The court did not address enterprise AI tools with negotiated confidentiality and no-training terms, but its focus on the missing confidentiality duty suggests the terms governing an AI tool will be significant in future cases.
Prompts and chat logs in discovery
Parties may serve requests for prompt histories, outputs, account tiers, platform terms, and AI use policies in future trade secret cases. Such evidence may show what was disclosed, to whom, and whether secrecy measures were adequate. It may also be relevant to knowledge and intent. Organizations may wish to address AI records in litigation holds and electronically stored information (ESI) protocols.
Privilege may not protect AI conversations
In United States v. Heppner (S.D.N.Y. Feb. 17, 2026), Judge Jed S. Rakoff of the Southern District of New York held that documents a criminal defendant created with a public AI tool were neither privileged nor work product, even though later shared with his attorney. Among other factors in its attorney-client privilege analysis, the court considered that the platform’s privacy policy permitted it to collect and use inputs, which weighed against a reasonable expectation of confidentiality. The court rejected work-product protection on separate grounds.
The law on AI and privilege is still developing. Legal analysis drafted in a public AI tool may face challenges to privilege and work-product protection.
Key considerations
- Contract for confidentiality. Confine sensitive work to enterprise AI tools with confidentiality, no-training, and limited-retention terms.
- Address shadow AI. Consider the uses and drawbacks of personal public AI accounts for company work.
- Write it down. Adopt a policy on what may be entered into AI tools. Train employees and document the training.
- Update agreements and protocols. Cover AI inputs in non-disclosure agreements, vendor agreements, litigation holds, and ESI protocols.
- Review before asserting claims. Confirm whether an asserted secret passed through a public AI tool.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.
[View Source]
