The Invisible Leak: Generative AI and the Threat to Corporate Trade Secrets

The adoption of generative AI has fundamentally shifted how businesses operate. Millions of professionals feed data into AI chatbots every day to draft communications, optimize software, or analyze complex financial records.

However, this reliance introduces a massive vulnerability to corporate intellectual property. When employees paste proprietary data or internal strategies into an AI prompt, they risk committing an irreversible legal misstep – the public disclosure and subsequent loss of their company’s trade secrets.

Understanding the intersection of confidential information and AI requires exploring three core challenges, including maintaining secrecy, the mechanics of trade secret misappropriation, and the realities of legal enforcement.

The Secrecy Requirement: Feeding the Machine

Under global frameworks – such as the U.S. Defend Trade Secrets Act (DTSA), the Uniform Trade Secrets Act (UTSA), and the EU Trade Secrets Directive – information only qualifies as a trade secret if it meets two strict conditions:

1. It derives independent economic value from not being generally known.

2. It is subject to reasonable efforts under the circumstances to maintain its secrecy.

Generative AI directly threatens the second condition. When a user enters text into a public generative AI platform, that prompt data is frequently processed, stored, and used to further train and refine future iterations of the model.

Legal Precedents

Recent court cases have underscored that the law will not protect companies that fail to police their AI usage.

In Trinidad v. OpenAI,¹ a district court in the Northern District of California dismissed a plaintiff’s trade secret claims under the DTSA. The court ruled that because the plaintiff voluntarily uploaded her proprietary frameworks directly into ChatGPT, she failed to take the legally required reasonable measures to protect their secrecy.

Similarly, in United States v. Heppner,² a court held that documents drafted using a public AI platform lacked the confidentiality required to claim attorney-client privilege. The court emphasized that users of standard public AI tools do not possess a reasonable expectation of privacy. In short, pasting a trade secret into an unsecure, public AI prompt is legally equivalent to publishing it on a public forum. Once the data enters the public domain, its status as a protectable trade secret is permanently extinguished.

The Court is aware that some commentators have argued that whether Claude is an attorney is irrelevant because a user’s AI inputs, rather than being communications, are more akin to the use of other Internet-based software, such as cloud-based word processing applications. But the use of such applications is not intrinsically privileged in any case, and the argument that Claude is like any other form of software only cuts against the invocation of privilege because all ‘[r]ecognized privileges’ require, among other things, ‘a trusting human relationship,’ such as, in the attorney-client context, a relationship ‘with a licensed professional who owes fiduciary duties and is subject to discipline.’“

Rakoff, J., United States v. Heppner

Misappropriation and Theft: Corporate Espionage via Prompt

Trade secret misappropriation occurs when a protected asset is acquired, used, or disclosed through improper means, or in breach of a duty of confidentiality. In an AI-dominated environment, misappropriation happens on two main fronts.

Inadvertent Employee Disclosure

Most AI-related data leaks are not malicious; they are accidental. An engineer trying to debug a proprietary software algorithm or an executive summarizing a pre-IPO financial forecast might paste sensitive code or text into an AI tool to save time.

If that model later exposes components of that information to a competitor during a related prompt query, a massive leak has occurred. While the employee may not have intended to harm the company, the business’s competitive edge is compromised, and pursuing the AI developer for trade secret theft becomes an uphill battle if the software’s standard terms of use permitted data ingestion.

Exfiltration and Prompt Injection

Conversely, bad actors are actively weaponizing AI to extract confidential corporate information. Competitors or cybercriminals can use sophisticated prompt injection attacks—crafting specific inputs designed to bypass an AI system’s built-in guardrails—to force a deployment model to output proprietary methodologies or underlying training data. Courts have affirmed that using automated bots or deceptive inputs to scrape or extract secret, economically valuable data from platforms constitutes acquisition by improper means.

Enforcement and Compliance: How to Protect Your Assets

Because an AI leak can instantly dissolve a company’s legal recourse, enforcement strategies must pivot from reactive litigation to aggressive internal prevention.

Implement Enterprise-Grade Gatekeeping

The most effective way to enforce trade secret boundaries is to control the infrastructure. Organizations are moving away from blanket prohibitions – which employees frequently circumvent – and are instead funneling activity through secure, contractually protected channels, including:

Opt-Out Provisions & Enterprise Licensing: companies must ensure employees only use enterprise tiers of AI tools. These specific corporate contracts explicitly guarantee that prompt data will never be used for model training, retained indefinitely, or shared with third parties.

API Ingestion: utilizing private API connections to interface with AI models generally bypasses the standard data-scraping terms applied to public consumer interfaces.

Update Internal Policies and NDAs

Non-disclosure agreements (NDAs) and employment agreements drafted before the generative AI boom are no longer sufficient. Legal teams are systematically revising these documents to explicitly state that inputting company-owned text, code, or customer lists into unauthorized third-party AI platforms constitutes a material breach of confidentiality and a misappropriation of corporate assets.

Takeaway

Generative AI is a powerful operational multiplier, but it operates as an information vacuum. In trade secret law, secrecy is an absolute, unforgiving standard. Once an asset is shared with an unsecured machine, the secret is out – and the legal protection for the corporate secrets vanishes with it. To survive this shifting landscape, modern businesses must treat every AI prompt box not as a private assistant, but as a public megaphone.

This blog posting is for informational purposes only. If you have a specific issue or question related to trade secrets or AI, please contact Yonaxis I.P. Law Group.

---

Footnotes

1. See Order Granting Motion to Dismiss First Amended Complaint, Case No. 4:25-cv-06328 (N.D. Cal. Jan. 5, 2026) (Tigar, J.). ↩

2. See Memorandum, Case No. 25 Cr. 503 (JSR), 2026 BL 52143, U.S. Dist. LEXIS 32697 (S.D.N.Y. Feb. 17, 2026). ↩