DOJ indicts former Google engineer for AI theft

The US Department of Justice indicted a former Google senior engineer this week for stealing proprietary AI training infrastructure designs and selling them to a Chinese AI startup.

The indictment alleges that the engineer, who worked on Google’s internal AI training infrastructure for 4 years, downloaded over 50,000 confidential files in his final weeks at Google. The files included:

• Internal AI accelerator chip designs.
• Data centre cooling and power designs.
• AI model training runbooks.
• AI training data catalogues.

The engineer allegedly transferred the files to a personal cloud storage account, then to a Chinese AI startup where he took a senior role.

The indictment charges include:

• Theft of trade secrets (18 USC § 1832).
• Economic espionage (18 USC § 1831).
• Transport of stolen property across state lines.

The maximum sentence is 15 years per count.

Why this matters:

The indictment is the highest-profile AI trade secrets case to date. Previous cases involved lower-level engineers or less sensitive technology.

The case highlights the strategic value of AI infrastructure IP. Google’s internal designs for AI training are arguably the company’s most valuable trade secrets. The fact that they’re being prosecuted as a federal case shows the US government’s view that AI IP is critical national infrastructure.

The China dimension is significant. The indictment alleges that the recipient was a Chinese AI startup, and the engineer’s role there was to “build out their AI training capability””. This matches a pattern of US-China tech transfer concerns that have escalated over the past 5 years.

The legal context:

Trade secrets cases are hard to prosecute. The government must prove:

• The information was actually a trade secret (not publicly available).
• The defendant knew it was a trade secret.
• The defendant took it without authorisation.
• The defendant intended to use or benefit from it.

The case has documentary evidence (the engineer admitted in internal messages to “backing up” the files). The prosecution has a strong case.

The Chinese destination adds complexity. If the engineer transferred the files while physically in China, US jurisdiction is unclear. The indictment doesn’t specify where the transfer occurred.

Implications:

For US AI companies, the case is a reminder that trade secret protections are critical. AI infrastructure designs, training data catalogues, and model training recipes are among the most valuable trade secrets in any company.

The case is likely to accelerate:

• Stricter employee offboarding processes at AI companies.
• Increased use of digital rights management (DRM) on AI infrastructure files.
• More aggressive prosecution of AI trade secret cases.
• Increased scrutiny of AI talent movements to China.

The case also signals that the US government’s view of AI is changing. AI is no longer just a commercial sector — it’s strategic infrastructure. The trade secrets laws are being applied accordingly.