Employee Departures and Trade Secret Risk in the AI Era

On August 28, 2025, xAI (formerly known as Twitter) filed suit in the Northern District of California against a former employee, alleging willful misappropriation of its most sensitive trade secrets, breach of contract, fraud, and violations of California’s computer access statutes. X.AI Corp. et al v. Xuechen Li, Case No. 3:25-cv-07292 (N.D. Cal. Aug. 28, 2025). The allegations are nothing short of explosive.

According to the complaint, the employee copied highly confidential materials relating to xAI’s Grok model—information that the company describes as the culmination of billions of dollars of investment and years of engineering—and then concealed that theft (in addition to cashing out millions of dollars worth of stock) before resigning to join OpenAI, one of xAI’s direct competitors.

The complaint alleges that among the trade secrets that the defendant misappropriated, xAI’s “cutting-edge AI technologies with features superior to those offered by ChatGPT and other competing projects” were included. The complaint further alleges that the defendant actually admitted to misappropriating confidential information and trade secrets, and that he admitted to attempting to “hide his theft.”

A central fact in the complaint is the role of a “termination certification.” Like many companies, xAI required departing employees to sign a statement confirming that they had returned all confidential information, deleted any copies from personal devices, and would continue to protect the company’s secrets. The complaint alleges that the employee signed such a certification just days after transferring sensitive files to personal accounts, and details each of the misrepresentations made by the employee in the certification.

ourts have long viewed written acknowledgments of confidentiality obligations (including exit certifications, NDAs, intellectual property agreements, etc.) as evidence that a company took reasonable measures to protect its trade secrets.^[1] These acknowledgments reinforce an employee’s obligations, create a record that the company sought to preserve secrecy, and provide support for remedies if litigation becomes necessary. Generally, if an employer can show that the employee expressly recognized the ongoing duty of confidentiality, courts are more likely to find that the company satisfied its obligations under trade secret law^[2].

But as the xAI complaint illustrates, exit certifications are not foolproof. A determined employee can misrepresent compliance, leaving the company vulnerable despite its contractual and procedural protections. Their value lies in showing that the company took appropriate, documented steps – not in eliminating risk altogether. In this case, the misconduct was discovered through another layer of the protection program: routine reviews of logs from security software, designed to detect data exfiltration.

In the AI industry, the stakes are particularly high. Trade secrets include model weights, training data, system prompts, and tuning methods. These assets are both expensive to create and highly portable. Training a frontier model can cost hundreds of millions of dollars, and a single upload to personal cloud storage can compromise years of work. This combination makes employee departures a flashpoint for misappropriation claims. Companies that rely on trade secret protection must be prepared to demonstrate not only the presence of robust contractual restrictions, but also the technical and procedural steps they use to enforce them.

The xAI case underscores several lessons. Exit certifications remain useful, but they should be viewed as part of a broader offboarding process rather than a stand-alone safeguard. Technical measures such as monitoring for unusual access patterns or large data transfers can provide early warning of problems. Forensic readiness is also essential, since the ability to preserve logs and audit trails makes it far easier to detect misconduct and present credible evidence in litigation.

As I explained in a previous blog post, trade secret law is already shifting under the pressure of AI. Courts may begin to define what is “readily ascertainable” not by what a human could compile, but by what AI can recreate in seconds. Likewise, the “reasonable measures” requirement will flex to reflect these new risks. Exit certifications and traditional safeguards will remain important, but companies should also be prepared to demonstrate that their protective measures account for the new capabilities of AI itself.

The allegations in the complaint are a reminder that in the AI era, protecting trade secrets requires layered precautions. Contracts and exit certifications continue to matter, but they must be backed by technical controls, monitoring, and a culture of confidentiality. Courts will continue to ask whether reasonable measures were taken, but the answer to that question will increasingly be judged against the realities of many industries where the most valuable assets can be taken in seconds.

Sarah Tishler is the author of this article. Sarah is a commercial litigator whose practice focuses on complex business disputes in state and federal courts, including intellectual property, breach of contract, and fraud claims. Her experience also includes regulatory investigations, international arbitration, securities litigation, shareholder litigation, class-action litigation, and pre-transactional advising.