Background
Linwei Ding was a software engineer at Google who worked on the company’s AI computing infrastructure — specifically the systems that train large-scale machine learning models. Between May 2022 and April 2023, Ding secretly uploaded more than 2,000 pages of Google’s confidential technical documents to a personal Google Cloud account, including specifications for Google’s custom AI chips (TPUs) and the proprietary software stack used to operate them. Ding simultaneously began consulting for two AI startups in China.
Federal prosecutors charged Ding with seven counts of trade secret theft and seven counts of economic espionage — the more serious charge, which applies when trade secrets are stolen to benefit a foreign government or company. After an 11-day jury trial, Ding was convicted on all 14 counts in January 2026. He then moved for a new trial, arguing that the government had engaged in improper racial discrimination during jury selection by systematically striking jurors of Chinese descent in violation of the Supreme Court’s Batson v. Kentucky framework.
The Court’s Holding
Judge Chhabria denied the motion for a new trial on June 15, 2026. The court found that Ding had not established that prosecutors exercised their peremptory strikes in a racially discriminatory manner. Under Batson, a defendant must first make a prima facie showing of racial discrimination in juror exclusions; the burden then shifts to the government to offer a race-neutral explanation; and the court must then decide whether the explanation is credible or pretextual.
Judge Chhabria concluded that the government offered sufficient race-neutral justifications for its strikes and that Ding had not demonstrated that those explanations were a pretext for discriminatory exclusion. The jury that convicted Ding was properly constituted, and no basis for a new trial was established.
Ding remains convicted on all 14 counts. Sentencing was pending as of the date of this ruling.
Key Takeaways
- Criminal trade secret theft prosecutions under the Economic Espionage Act (18 U.S.C. §§ 1831–1832) can succeed even without direct evidence of disclosure to a foreign entity, as long as the government can show the thief intended to benefit foreign parties.
- Uploading proprietary technical documents to personal cloud storage — even without transmitting them — can constitute trade secret theft. The act of misappropriation is complete when the trade secret leaves the employer’s control.
- Jury-selection challenges under Batson are difficult to win post-verdict; a defendant must make a specific factual showing of discriminatory intent, not merely a statistical inference from the demographics of struck jurors.
- This case is a high-profile illustration of the government’s increased focus on protecting U.S. AI technology from foreign economic espionage — particularly involving technology developed for AI training infrastructure.
Why It Matters
The Ding prosecution is one of the most significant AI trade-secret criminal cases to date. The stolen materials related to Google’s Tensor Processing Units (TPUs) and associated training software — infrastructure that represents billions of dollars of investment and gives Google a competitive advantage in AI development. The government’s theory was that Ding planned to use this knowledge to benefit Chinese AI startups, raising the economic-espionage charge that carries penalties far beyond ordinary theft.
For technology companies, the case demonstrates the importance of access controls and monitoring on proprietary technical systems: Ding was able to exfiltrate thousands of documents without triggering alarms. It also shows that criminal prosecution — not just civil litigation — is increasingly available for trade-secret theft involving AI technology, particularly when foreign state-linked entities are implicated. The denial of a new trial means Ding’s convictions stand, and a lengthy sentence is likely on all 14 counts.
Surfaced via newsletter intake: Law360 IP.