Ex-Google engineer convicted of stealing AI secrets for Chinese firms

A former Google engineer was convicted in San Francisco on January 29, 2026 of stealing AI trade secrets to benefit two Chinese companies, according to the US Department of Justice.

Linwei Ding, a 38-year-old Chinese national, was found guilty after an 11-day trial of seven counts of economic espionage and seven counts of trade secret theft, involving thousands of pages of confidential information.

Each economic espionage charge carries a maximum penalty of 15 years in prison and a US$5 million fine, while each trade secret charge can result in up to 10 years and a US$250,000 fine.

Ding is scheduled for a status conference on February 3.

Prosecutors said Ding stole details about Google’s hardware infrastructure and AI training platform, including chip blueprints intended to reduce reliance on Nvidia and improve Google’s competitive edge.

Ding joined Google in May 2019 and began thefts in 2022 amid interest from Chinese tech firms, the DOJ said.

🔗 Source: Reuters

🧠 Food for thought

Implications, context, and why it matters.

Key details on the stolen AI technology and Ding’s Chinese-company ties remain undisclosed

Specific Google assets, including exact chip layouts or software source code, have not been publicly described in the Reuters report or the Justice Department press release.
The stolen trade secrets were described as detailed information on the architecture and operation of Google’s Tensor Processing Unit (TPU) chips (custom AI accelerators), Google’s Graphics Processing Unit (GPU) systems (widely used processors for AI workloads), software used for chip-to-chip communication and task execution, and software that coordinates thousands of chips into a supercomputer, plus trade secrets tied to Google’s custom-designed SmartNIC (a network interface card with onboard computing used to speed up and secure data-center networking) 1.
A secret affiliation with two PRC-based technology companies was alleged, with PRC referring to the People’s Republic of China, including one where chief technology officer talks were said to have occurred, and a later PRC-based AI and machine-learning company was said to have been founded and run as CEO; the companies were not named in the Justice Department press release and Reuters did not name them 1.
Those gaps are needed for the competitive impact on Google’s AI infrastructure and hardware roadmap to be sized.

The conviction puts a spotlight on insider safeguards for AI and chip IP

Security vendors and investors may see more spending on tools that keep AI and semiconductor intellectual property (IP) safe from insider theft.
The Justice Department said that from about May 2022 to May 2023, Ding uploaded more than 1,000 unique files with Google confidential information from Google’s network to his personal Google Cloud account 1.
Reuters reported that Ding was convicted after an 11-day trial on seven counts of economic espionage plus seven counts of theft of trade secrets, with coordination through the interagency Disruptive Technology Strike Force created in 2023 1.
That focus could lift interest in protections for chip design files, plus access controls for AI code and model repositories.