When Winning the Battle Risks Losing the War: xAI’s Trade Secret Case Against OpenAI Dismissed

In trade secret litigation, injunctive relief can be a powerful weapon. But xAI Corp. v. OpenAI, Inc. illustrates how an early courtroom win can undermine a longer-term corporate liability theory.

On February 24, 2026, the Honorable Rita F. Lin of the Northern District of California dismissed xAI’s trade secret misappropriation claims against OpenAI, finding that the complaint was “notably absent” of allegations about OpenAI’s own conduct. The ruling underscores that no matter how egregious an employee’s conduct, a trade secret claim under the Defend Trade Secrets Act (“DTSA”) requires allegations that the defendant acquired, induced, or used the stolen information, not just that it hired the people who took it.

The Facts

In the summer of 2025, eight xAI employees left for OpenAI in rapid succession. xAI’s First Amended Complaint (“FAC”) painted a picture of a coordinated raiding campaign, led by a single OpenAI recruiter named Tifa Chen, who communicated with multiple departing employees via the encrypted messaging app Signal. The most dramatic allegations involved two engineers: Xuechen Li and Jimmy Fraiture.

Li, one of xAI’s first 20 engineers, allegedly uploaded xAI’s entire source code base to a personal cloud account on July 25, 2025. He then deleted his browser history and system logs to cover his tracks. He later confessed in a handwritten statement, prepared with counsel present, to taking the code and a confidential slide deck on xAI’s model training and tuning methods. Fraiture, the second engineering hire in xAI’s London office, allegedly used AirDrop to copy source code, infrastructure and training code, co-founders’ experimental folders, and a recording of an internal all-hands meeting where Elon Musk discussed xAI’s product roadmap and commercial contract terms. xAI Corp. v. OpenAI, Inc., No. 3:25-cv-08133-RFL, FAC ¶¶ 59, 66, 98–102 (N.D. Cal. Oct. 27, 2025).

The FAC also identified an unnamed senior finance executive who, after only a few months at xAI, left for a data-center-focused role at OpenAI. He had no data center experience before joining xAI, and he himself had acknowledged that xAI’s data center operations were the company’s “secret sauce.” When xAI’s outside counsel, Alex Spiro, reached out about confidentiality concerns, the executive replied: “S__ my d___.” (One can read the Amended Complaint for the full phrase if unsure.) When xAI’s HR team followed up, he responded: “I quit. Leave me the f__ alone.” Id. ¶¶ 114, 118–19.

Other former employees allegedly retained work chats on personal devices, shared xAI information with an unidentified third party while employed by OpenAI and represented by OpenAI’s counsel, and, in the case of one engineer, Ruddarraju, attempted to access an internal xAI document about data center optimization and hiring after he had already started working at OpenAI. xAI blocked the attempt. Id. ¶¶ 127–30.

The TRO That Cut Both Ways

Before filing the corporate case against OpenAI, xAI moved against Li individually. In xAI Corp. v. Li, No. 3:25-cv-07292-RFL (N.D. Cal.), Judge Lin granted a temporary restraining order on September 2, 2025, barring Li “from having any role or responsibility at OpenAI … pertaining to generative AI including without limitation OpenAI’s ChatGPT until xAI has confirmed that all of xAI’s Confidential Information in Li’s possession, custody, or control has been deleted.” The order required Li to surrender his personal devices and cloud accounts for forensic review within three business days.

The TRO worked. OpenAI revoked Li’s job offer and he never worked at OpenAI.

But that success created a problem for the corporate case. Because Li never worked at OpenAI, xAI could not allege that OpenAI ever used any trade secrets transferred to it by Li. The TRO that kept Li out of OpenAI also kept xAI from arguing that OpenAI put his alleged misappropriation to use.

The Dismissal

Judge Lin’s 15-page order examines the allegations against each of the eight former employees and, one by one, finds them insufficient to state a claim against OpenAI.

On direct misappropriation, xAI offered no nonconclusory allegations that OpenAI itself acquired, disclosed, or used xAI’s trade secrets. The complaint relied on an inducement theory, but the court found no factual basis for that inference. Li and Fraiture transferred source code around the same time they were communicating with the same recruiter via encrypted messaging. But temporal proximity and encrypted communications, “[w]ithout more,” were “insufficient to support a plausible inference that OpenAI encouraged the alleged theft.” Slip op. at 8–9.

The court contrasted xAI’s allegations with cases where inducement was adequately pled. In Citcon USA, LLC v. RiverPay Inc., No. 18-cv-02585-NC, 2019 WL 917056 (N.D. Cal. Feb. 25, 2019), the defendant promised co-founder status conditioned on the employee retaining a device from the plaintiff. In Elko, Inc. v. WTH Commercial Services, LLC, No. 22-cv-00015-MMD, 2023 WL 6141623 (D. Nev. Sept. 20, 2023), the defendant entered into an agreement to steal confidential information and then used it to target the plaintiff’s customers. Nothing comparable was alleged here.

Any indirect misappropriation theory would require xAI to allege that OpenAI knew or had reason to know the information was a trade secret acquired by improper means. xAI alleged that Li disclosed confidential information during an interview presentation but did not allege facts permitting an inference that OpenAI knew the information constituted trade secrets or was improperly obtained. The connection between Li’s personal cloud account and his ChatGPT account, which, the complaint alleged, gave OpenAI the technical ability to “see and download” Li’s files, did not change the analysis. Slip op. at 9–10.

The vicarious liability theory was the heart of the case. Under respondeat superior, an employer can be vicariously liable for an employee’s misappropriation, but courts require allegations that the employer subsequently used the trade secrets or that the employee used them on the job. Alert Ent., Inc. v. Rana, No. 22-cv-06646-JSC, 2023 WL 2541353, at *3 (N.D. Cal. Mar. 16, 2023). The court found xAI’s allegations fell short on every front:

Li never worked at OpenAI. Fraiture did, but xAI did not allege he ever used the stolen code in his new job. To hold otherwise, the court reasoned, would automatically subject future employers to trade secret liability every time they hired someone who, without their knowledge, had taken confidential information from a prior employer on the way out the door. The court emphasized that “mere possession of trade secrets is not sufficient to constitute misappropriation,” citing Acculmage Diagnostics Corp. v. Terarecon, Inc., 260 F. Supp. 2d 941, 951 n.5 (N.D. Cal. 2003). Slip op. at 11.

The finance executive merely possessed data center knowledge, and the court noted that “[h]ostility toward one’s former employer during departure does not, without more, indicate use of trade secrets in a subsequent job.” Ruddarraju tried to access an internal xAI document but was blocked—so there was no basis to infer use. Pham and Knight shared information with a third party, but nothing in the FAC connected that disclosure to OpenAI’s interests or to anything they were hired to perform. Two other former xAI employees who had worked on xAI data center solutions or Grok 4 development (Pothanis and Dalton) simply left for OpenAI. xAI did not allege they took anything at all. Slip op. at 11–14.

The court also dispensed with xAI’s California Unfair Competition Law (“UCL”) claims, finding them preempted by the California Uniform Trade Secrets Act (“CUTSA”). xAI’s “unfair” prong claim was grounded in alleged poaching, but the court found that the poaching allegations “all focus on poaching in service of acquiring xAI’s trade secrets and do not identify any other reason why the hiring of those employees was anticompetitive.” Slip op. at 15.

Takeaway

The dismissal was with leave to amend; xAI has until March 17, 2026, to file a Second Amended Complaint. But Judge Lin’s roadmap is clear: to survive, xAI needs facts tying OpenAI to the acquisition or use of its trade secrets, not just facts about what departing employees did on the way out.

The order is also a case study in the tension between seeking early injunctive relief and building a corporate liability theory. xAI’s TRO against Li was an unqualified short-term success: it kept a confessed trade secret thief out of its chief rival’s offices. But it also ensured that the most damaging factual predicate (Li actually working at OpenAI) never materialized. For xAI, the benefits of keeping Li from working at OpenAI were valuable in and of themselves, but from a purely litigation perspective, the TRO success diminished xAI’s claims.