xAI, the artificial intelligence startup led by Elon Musk, filed a lawsuit in California federal court against former engineer Xuechen Li. The complaint alleges that Li, a Stanford-trained AI researcher, willfully copied proprietary data related to xAI’s Grok chatbot and transferred it to his personal computers, intending to deliver it to rival OpenAI upon joining their ranks.
This incident isn’t just a corporate dispute; it’s a stark illustration of a targeted insider threat exploiting privileged access to exfiltrate sensitive intellectual property (IP). The stakes are high: xAI claims the stolen data includes “cutting-edge AI technologies with features superior to those offered by ChatGPT,” positioning this as a significant breach in the hyper-competitive AI development landscape.
Xuechen Li joined xAI in February 2024 as one of its earliest engineers, working on the training and development of the Grok chatbot. His role granted him access to highly sensitive data, including proprietary algorithms and model architectures deemed superior to OpenAI’s ChatGPT. This privileged access was the foundation of the breach.
In June 2025, Li sold $4.7 million worth of xAI stock, followed by an additional $2.2 million in July, totaling $7 million. xAI facilitated these transactions to retain Li, indicating trust in his contributions. However, on July 25, 2025—the same day the second stock sale was finalized—Li allegedly copied confidential data from his xAI-issued laptop to personal storage systems.
To cover his tracks, he reportedly deleted browser history, system logs, and renamed or compressed files. These actions suggest premeditation and an intent to obscure forensic evidence.
The data transfer occurred just days before Li’s resignation on July 28, 2025. The complaint alleges that Li had already accepted a job offer from OpenAI, with a start date of August 19, 2025. The timing suggests that the exfiltration was strategically aligned with his departure and new role.
xAI’s security team identified the breach during a routine review of logs from data loss prevention (DLP) software. This software, likely configured to detect unauthorized data transfers, flagged anomalies tied to Li’s activities. On August 11, xAI emailed Li, demanding the return and deletion of the stolen data.
In a meeting with xAI’s legal team, Li, accompanied by a criminal attorney, reportedly admitted to taking the files and attempting to conceal his actions.
He provided a handwritten statement confirming the theft and allowed xAI to copy his personal laptops for forensic analysis. However, xAI alleges that Li withheld passwords for critical accounts, hindering a full investigation. Subsequent analysis revealed additional undisclosed stolen data on his devices.
The xAI lawsuit does not name OpenAI as a defendant, suggesting no direct evidence that OpenAI orchestrated the theft. However, the timing of Li’s job offer and data exfiltration raises questions about competitive espionage.
