X.AI Suit Brings Trade Secrets Litigation to Fertile Ground

The trade secrets lawsuit X.AI filed against a former engineer heralds an aggressive approach from generative AI companies in a landscape driven by exploding revenues and a competitive hunt for top talent.

Accomplished AI architect Xuechen Li is accused by his former company of stealing “entire code base” and liquidating nearly $7 million in employee stock before joining competitor OpenAI. Before the civil case was filed, the FBI opened a criminal investigation into the alleged data theft, according to a court filing by Li’s attorney.

While the basic pattern is typical in trade secrets cases against ex-employees, the alleged theft of generative AI data in an industry swiftly boasting multi-billion dollar valuations sets it apart. The rapidly advancing technology and hyper-competitive race to hire—or poach—a finite supply of elite talent creates a perfect storm over the safety of trade secrets.

Companies will respond by tightening security and moving quickly against suspect violations, Craig Smith of Lando & Anastasi LLP said. X.AI’s rapid response to Li’s alleged theft—it confronted him just two weeks after his departure and sued two weeks later—underscores the mechanisms in place for detecting and quickly acting, Smith said.

“The Li case is a signal to the industry that these companies are going to be very aggressive when it comes to protecting their trade secrets and monitoring the activities of their employees,” Smith said. “When you have this unprecedented competition for talent in the AI space, you’re going to see companies be much more aggressive in how they enforce their trade secrets.”

The FBI executed a search warrant on Li on Aug. 19, his attorney said during the Sept. 2 hearing. That indicated the company tipped off authorities, according to attorney Fred Norton of The Norton Law Firm, who said he expects other companies will respond by beefing up their efforts to protect their secrets.

“In-house counsel, information security people are going to be more careful about how tech is locked down and pay attention to departing employees,” Norton said. “There are going to be lawsuits in any area where you’ve got rapidly developing technology and market value.”

The Sept. 2 hearing led to a temporary restraining order forcing Li to return any X.AI materials, and barring him from working for or communicating with OpenAI or any competitor until the information was deleted. There will be a hearing on X.AI’s bid for a preliminary injunction Oct. 7.

Siloed Sophistication

Trade secrets are intellectual property that aren’t publicly disclosed—unlike patents, trademarks, and copyrights—and inherently derive value from confidentiality. The importance for companies to design processes to protect secrets has only grown, attorney Kelly Lawton-Abbott of SSM Law PC said, pointing to common practices like confidentiality agreements, employee education, limiting access to confidential information on a need-to-know basis, and tracking electronic access downloads.

Against that backdrop, she said she found the details in X.AI’s suit surprising.

If true, “it’s interesting that someone on that level of tech sophistication thought that would work,” said Lawton-Abbott, who focuses on helping tech companies protect their proprietary information.

Norton said what is alleged happens more than one might expect.

“The reality is the sophistication of any of these people is kind of siloed,” Norton said.

It’s possible to steal trade secrets even without risking electronic detection, JD Harriman of Foundation Law Group added.

Employees aren’t required to functionally forget everything they ever learned, Harriman said, but concerted efforts to memorize proprietary information may be actionable—if much harder to prove.

‘A Good Lesson’

Harriman said he found it unsurprising a generative AI case hadn’t yet arisen. Most workers know not to steal trade secrets and companies are good at protecting them, he said. On top of that, AI programs require vast amounts of memory, making them difficult to take wholesale.

But more accessible segments of potentially valuable code and other information do exist—and the stakes are rising.

“Anytime there are huge amounts of money getting thrown at something and it’s in the growth phase, that’s where it’s most likely that trade secrets get stolen,” Harriman said.

Generative AI companies are offering enormous financial gains in the form of nine-figure packages and even $100 million signing bonuses to secure top talent. Smith called the “arms race for AI talent” and resulting salaries “incredible” and said the situation can be tricky for the employees, despite strong incentives to avoid litigation and reputational damage.

“A lot of them are young and may have no experience with trade secrets law or best practices for leaving one company and joining another,” he said.

Norton, who used to work for Tesla and whose law firm formerly represented Twitter, said there’s no way to make it impossible to take secrets while still letting employees do their jobs. But education, vigilance when employees leave, and exit interviews can head off a lot of problems, he said.

Lawton-Abbott pointed to the importance of companies making clear to employees what, exactly, their trade secrets are. Harriman added that companies should carefully review language in their confidentiality agreements for oversights such as failing to specify that promises to safeguard trade secrets don’t expire alongside other aspects of non-disclosure agreements.

“If someone says you have to keep this secret until Thursday, that means on Friday you don’t have to keep the secret,” he said.

Norton also said companies poaching employees should be cautious to avoid being roped into lawsuits—OpenAI, for example, isn’t a defendant in X.AI’s suit. He advised hiring a third-party attorney to provide new employees guidance and take information the employee shouldn’t have so the hiring company can say it never had access.

“It’s a good lesson for companies of all size about having safeguards and termination procedures in place,” Lawton-Abbott said.

The case is X.AI Corp. v. Li, N.D. Cal., No. 3:25-cv-07292.