Trade Secrets and Public AI Tools: What Trinidad v. OpenAI Means for Texas Businesses
A federal court has dismissed a trade-secret claim built entirely inside a public AI tool, and the reasoning matters well beyond the case. The claimed secrets were developed through ChatGPT, which meant they were voluntarily shared with a company that owed no duty to keep them confidential. The claim failed on the secrecy element that every trade-secret claim, in Texas and under federal law, must prove.
What happened.
Rebecca Trinidad sued OpenAI in federal court, representing herself. She alleged that she had developed artificial intelligence "frameworks," which the complaint calls her "protocols and frameworks," through her use of ChatGPT, and that OpenAI took them and incorporated them into its products, including the ChatGPT Agent feature. Her claims included copyright infringement, misappropriation of trade secrets under the federal Defend Trade Secrets Act (DTSA), California unfair-competition claims, and unjust enrichment.
On January 5, 2026, Judge Jon S. Tigar of the Northern District of California dismissed the amended complaint with prejudice. The copyright claim failed because Trinidad had not registered any copyrights, a precondition to an infringement suit. The trade-secret claim failed at the threshold, on secrecy, and it is that part of the order that drew practitioners' attention.
Why the trade-secret claim failed.
A DTSA plaintiff must first show that the information was a trade secret. Under 18 U.S.C. § 1839(3), that means the owner "has taken reasonable measures to keep such information secret" and the information derives independent economic value from not being generally known. The site's guides explain that framework, including what counts as a trade secret in court. Trinidad's claim failed on the first element.
She alleged no measures to keep the frameworks secret. By her own account, she developed them using ChatGPT. The court agreed with OpenAI that developing them there "would have required her to voluntarily share the information she now alleges is part of her 'trade secrets' with OpenAI." Quoting Ruckelshaus v. Monsanto Co., 467 U.S. 986, 1002 (1984), the court held that a person who discloses a trade secret "to others who are under no obligation to protect the confidentiality of the information" sees the "property right is extinguished."
Two arguments did not change the result. First, ownership: Trinidad pointed to OpenAI's Terms of Use, under which she said she owned the outputs of her conversations. The court answered that ownership is not the test. She had to show the information was "secret" within the meaning of § 1839(3), and "whether she owned those ideas notwithstanding the disclosure is not relevant." Second, consent: she argued that any consent in the Terms was part of an unenforceable contract of adhesion. The court held that she had consented to the disclosure by accepting the Terms, and that the disclosure stood "whether or not [her] consent is enforceable as a contractual matter," because the DTSA asks what measures the owner took, not whether the disclosure was fair.
The limits of the ruling.
Trinidad is a single district-court order deciding a motion to dismiss, in a case brought by a self-represented plaintiff on unusual allegations. It binds no other court, and the facts are far from a documented business dispute between companies. Its significance is narrower and more practical: the court applied the ordinary DTSA framework, the same reasonable-measures inquiry Texas courts apply under the Texas Uniform Trade Secrets Act, which likewise defines a trade secret by the owner's efforts that are reasonable under the circumstances to maintain secrecy (Tex. Civ. Prac. & Rem. Code § 134A.002). Practitioner alerts, including one from DLA Piper, flagged the case for that reason. Nothing in the analysis depended on exotic features of AI: information handed to a third party outside the owner's confidentiality arrangements is information the owner will have difficulty calling secret, whoever or whatever the third party is.
A related split over AI and confidentiality.
Trade-secret secrecy is not the only doctrine asking what happens when information passes through an AI platform. In February 2026, two federal courts answered a related question about litigation privilege in opposite ways. In United States v. Heppner (S.D.N.Y.), Judge Rakoff held that documents a criminal defendant generated with Anthropic's Claude were protected by neither attorney-client privilege nor the work-product doctrine: Claude is not an attorney, and the consumer tool's terms exposed the material to the provider (bench ruling Feb. 10, 2026; written opinion Feb. 17, 2026). The same day, in Warner v. Gilbarco Inc. (E.D. Mich.), a magistrate judge held that a self-represented plaintiff's ChatGPT materials were protected work product and that using the tool did not waive that protection. The doctrines and facts differ, but the shared point matters: this area of law is unsettled, and the tier and terms of the tool are part of the analysis.
Practical points for Texas businesses.
- Public tools sit outside the confidentiality perimeter. Information entered into a consumer AI tool is shared with the provider under the provider's standard terms. If that information is later claimed as a trade secret, the sharing is the first fact a court will examine.
- Enterprise terms are the control. Business tiers in which the provider commits not to train on customer inputs, and accepts confidentiality and data-processing obligations, put the provider in a position closer to a vendor under a nondisclosure agreement. The disclosure analysis follows the obligations.
- Policies should name the tools. Guidance that distinguishes public tools from approved enterprise tools, and reserves public tools for work that is not sensitive, is itself evidence of reasonable measures. The guide to trade secret law in Texas collects the measures Texas courts credit.
- Secrecy is judged on what the owner did. Under both the DTSA and Texas law, the reasonable-measures element examines the owner's own conduct: access limits, confidentiality agreements, and channel discipline, which now includes which AI tools the business permits for proprietary work.
What to watch.
Whether other courts adopt Trinidad's threshold reasoning for claims built on work done in consumer AI tools, and whether any court distinguishes enterprise deployments backed by no-training and confidentiality commitments. How Texas courts apply the state statute's reasonable-measures requirement when the alleged disclosure is to an AI provider. And whether an appellate court resolves the Heppner and Gilbarco split over privilege and work product for AI-assisted work.
Sources.
- Trinidad v. OpenAI Inc., No. 25-cv-06328-JST (N.D. Cal. Jan. 5, 2026) (order granting motion to dismiss first amended complaint, with prejudice). law.justia.com (retrieved October 20, 2026).
- DLA Piper, "Trinidad v. OpenAI: Court Addresses Trade Secret Protection And Use Of Public AI Tools" (2026). mondaq.com; also at jdsupra.com (retrieved October 20, 2026).
- Kirkland & Ellis, "Two Federal Courts Chart Diverging Paths on the Discoverability of LLM Interactions" (2026), on United States v. Heppner, 2026 WL 436479 (S.D.N.Y. Feb. 17, 2026), and Warner v. Gilbarco Inc. (E.D. Mich. Feb. 10, 2026). kirkland.com (retrieved October 20, 2026).
- Beck Reed Riden LLP, "Who owns an AI-generated trade secret?" beckreedriden.com (retrieved October 20, 2026).
- Ruckelshaus v. Monsanto Co., 467 U.S. 986, 1002 (1984), as quoted and applied in the Trinidad order.
This article is general information, not legal advice. Law checked through October 20, 2026. See the Disclaimer.