AI in litigation: AI can extinguish your trade secret

November 19, 2026

The rapid adoption of generative artificial intelligence (GenAI) tools has introduced a host of potential legal risks. As courts and practitioners work through the implications of GenAI use, much of the conversation has focused on whether the use of GenAI waives attorney-client privilege or work product protections. But companies should be aware that inputting information into GenAI has implications for other protections as well.

For example, companies should be aware that GenAI usage can compromise a company’s trade secrets. Specifically, recent court decisions have found that, under certain circumstances, inputting proprietary information into a GenAI platform immediately extinguishes a company’s trade secret rights in that information. And, unlike privilege waiver (which can sometimes be cured or cabined within the context of litigation), trade secret waiver has no such safety valve—once the secret is extinguished, the property right is gone. Companies must be aware of this issue and take reasonable steps to protect against it.

In this post, we discuss a recent federal court decision, Trinidad v. OpenAI, that addressed the consequences of a party disclosing trade secrets to a GenAI platform. There, the court found that the disclosure of the trade secret to the GenAI platform—by inputting the trade secret into the GenAI tool—was fatal to the trade secret. This post also provides key takeaways for companies looking to safeguard their trade secrets in the world of AI.

Trade secrets play by different rules than privilege

We have previously written about the effect of inputting privileged information into GenAI platforms.1 While useful in the context of privilege, companies should not assume those decisions apply similarly to trade secrets. Unlike privilege, which is a protection that can attach and detach depending on context, confidentiality and the relationship between the parties, trade secret protection is generally tied to the owner’s ongoing obligation to protect the secrecy of the information. A trade secret that is no longer secret is, by definition, no longer a trade secret.2 There is no analogue to inadvertent disclosure, no “clawback” mechanism like those used in litigation to protect privileged information and no protective order that can restore the status quo once that disclosure has been made.

This distinction matters for companies using AI tools in everyday business operations as well as those whose only interaction with AI is via employees’ personal use. No matter how a company uses AI, if an employee inputs proprietary methodologies, business processes, customer data, pricing strategies or other competitively sensitive information into a publicly available GenAI platform, Trinidad v. OpenAI should put the company on notice that it may be forfeiting its trade secret rights in that information at the moment of input.

Trinidad v. OpenAI

In Trinidad v. OpenAI Inc., the Northern District of California dismissed a pro se plaintiff’s trade secret misappropriation claim under the Defend Trade Secrets Act (DTSA) for this specific reason. It found, based on the pleadings and briefing, that the plaintiff had voluntarily disclosed to ChatGPT the very information she claimed as a trade secret.3 The court found that her voluntary disclosure to the GenAI platform extinguished her trade secret, because her disclosure did not involve reasonable measures to keep the information secret and she disclosed it to others (OpenAI, the company that operates ChatGPT) that were under no obligation to protect the confidentiality of her trade secrets.

As background, the plaintiff brought an action against OpenAI, alleging that the company had “unlawfully appropriated her ‘proprietary methodologies for cultivating emergent identity and sovereign agency in artificial intelligence systems.’”4

To succeed on a DTSA claim, a plaintiff must show, among other things, that she possessed a trade secret, which is information that the owner has taken “reasonable measures to keep secret.”5 The court found the plaintiff had not alleged that she took any such measures. Rather, she admitted that she developed her “frameworks” using ChatGPT, which meant that she voluntarily shared the information she claimed as a trade secret with ChatGPT and, by extension, OpenAI.6

The court concluded that, because the plaintiff here “disclose[d] [her] trade secret to others who are under no obligation to protect the confidentiality of the information, ... [her] property right is extinguished.”7

The plaintiff argued that OpenAI’s terms of use established that she owned the output of her ChatGPT conversations. But the court found this irrelevant, holding that the relevant inquiry under the DTSA is not ownership, but secrecy: by developing her ideas through ChatGPT, the plaintiff voluntarily disclosed the existence of her ideas to OpenAI, regardless of whether she owned those ideas, and the fact that she may have owned those ideas did not rescue her claim.8

The plaintiff also argued that her acceptance of the OpenAI terms of service was unconscionable and coercive. The court rejected this argument, reasoning that the relevant inquiry for the trade secret claim was whether the plaintiff took reasonable measures to keep the information secret.9

Privilege analysis in litigation does not apply to trade secrets

In the privilege context, the fact that a litigant is proceeding pro se is often treated as legally significant, as the work product doctrine can be extended to protect a pro se plaintiff’s work product.10 In Trinidad, however, the plaintiff’s pro se status made no difference to the court’s trade secret analysis. Unlike privilege, the trade secret analysis turned not on a plaintiff’s legal sophistication or relationship to counsel, but on whether the information at issue was kept secret. The moment the plaintiff in Trinidad entered her “frameworks” into ChatGPT, the property right in that information was extinguished. The takeaway from Trinidad is clear: disclosing a purported trade secret to a publicly available AI platform can extinguish the trade secret immediately, no matter who is doing the disclosure.

Spreading beyond the Ninth Circuit

The Trinidad Court’s decision has also been cited outside the Ninth Circuit. For example, a court in another circuit cited Trinidad for the proposition that a plaintiff who discloses his trade secret to others who are under no obligation to protect the confidentiality of the information has thereby extinguished his property right in that information.11 This suggests that courts outside the Ninth Circuit are treating the decision as at least persuasive guidance on the question of trade secret waiver through AI disclosure.

Practical steps for companies

The Trinidad decision should be seen as a warning for companies that the risk involved with GenAI use is not confined to litigation or even privilege. Employees who input sensitive information into publicly available GenAI platforms can affect a company’s trade secrets regardless of whether the company is in active litigation. Although this is an evolving area of the law, companies should consider taking the steps below to protect themselves.

Limit employee GenAI use to enterprise AI tools for any work involving confidential or proprietary information

Many companies are implementing enterprise versions of common GenAI tools, which are supposed to be confidential to the company and do not permit third parties to monitor the content of prompts and outputs generated by these enterprise tools. Companies should consider prohibiting the use of non-enterprise GenAI platforms.

Review and update AI governance policies with specific attention to trade secret risk

Existing policies may adequately address GenAI privilege concerns but may still fail to address the distinct trade secret risk. Companies should review their current policies and consider whether they clearly identify the categories of information that constitute trade secrets and expressly prohibit inputting any such information into AI tools that have not been vetted and approved by legal or IT security teams.

Audit current AI tool usage across the company

Conducting an audit of which AI tools employees are currently using, for what purposes and with what types of information could allow companies to identify and remediate existing exposures before a trade secret is waived.

Ensure contractual protections with third-party AI providers are robust and documented

Companies should also review and consider what contractual assurances they have in place with any permitted GenAI providers that process confidential information, addressing security, retention and privacy issues.

Conclusion

Companies should be aware that the use of GenAI carries a risk of trade secret waiver that is distinct from the privilege waiver risks associated with litigation. While courts are developing context-sensitive approaches to the privilege question, the answer to the trade secret question appears to be more categorical at this time; under current case law, companies should assume that disclosing trade secrets to a publicly available AI platform likely extinguishes the property right in that trade secret and act accordingly.


Footnotes

2   A trade secret is information that derives independent economic value from not being generally known to others who could benefit from its disclosure or use, provided the owner takes reasonable efforts to maintain secrecy. See, e.g., Tex. Civ. Prac. & Rem. Code Ann. § 134A.002(6); Cal. Civ. Code § 3426.1(d).

3   No. 25-CV-06328-JST, 2026 WL 21791 (N.D. Cal. Jan. 5, 2026), appeal dismissed, No. 26-721, 2026 WL 1251786 (9th Cir. Apr. 6, 2026).

4   Id. at *1.

5   Id. at *4.

6   Id.

7  

Id. (quoting Ruckelshaus v. Monsanto Co., 467 U.S. 986, 1002 (1984) (“If an individual discloses his trade secret to others who are under no obligation to protect the confidentiality of the information, or otherwise publicly discloses the secret, his property right is extinguished.”)).

8   Id.

9   Id.

11   Ramos v. OpenAI, Inc., No. 6:25-CV-03281-MDH, 2026 WL 1556475, at *3 (W.D. Mo. June 2, 2026).