Former Google Engineer Convicted in Landmark AI Theft Case; Jury Finds Linwei Ding Guilty of Economic Espionage for China
A federal jury in San Francisco convicted Linwei Ding, a former Google software engineer, on 14 counts of trade-secret theft and economic espionage for allegedly stealing confidential AI technology and transferring it to China-based companies.
- Defendant: Linwei (Leon) Ding, 38, former Google software engineer; charged and tried with theft for alleged links to China-based firms (U.S. Department of Justice, Los Angeles Times).
- Verdict: Guilty on all 14 counts — seven counts of theft of trade secrets and seven counts of economic espionage; prosecution detailed thousands of pages of proprietary documents copied.
- Timeframe: Prosecutors say the improper copying occurred from May 2022 through April 2023 and involved sensitive materials on TPUs, GPUs and SmartNICs.
- Next steps: Ding was released pending sentencing; a status conference is set for February 3, 2026, and statutory penalties are substantial though sentencing will follow federal guidelines.
What prosecutors allege
Federal prosecutors say Ding joined Google on May 13, 2019 and worked on software to improve GPU efficiency for machine-learning workloads at Google and Google Cloud data centers. They allege he had access to confidential design and performance information for Google’s Tensor Processing Units (TPUs), Graphics Processing Units (GPUs), and SmartNIC network cards — central infrastructure for training and running advanced AI models (The Register, Fox Business).
The U.S. Department of Justice says Ding began copying sensitive internal documents in May 2022, transferring files to personal cloud accounts while attempting to evade Google security systems. Prosecutors presented logs showing large uploads and later downloads to a personal computer shortly before he resigned (Help Net Security, The Register).
Ties to Chinese firms
Trial evidence tied Ding to two China-based companies. Prosecutors say he was the chief technology officer of one firm and founded Shanghai Zhisuan Technology Co. Ltd., described as aiming to build AI supercomputers and custom ML chips. He allegedly worked with a firm called Rongshu and met prospective investors in China.
Prosecutors also introduced an application Ding made to a Chinese government “talent plan,” which investigators say seeks to recruit expertise to boost computing infrastructure. The government argued these links demonstrated intent to help China’s AI development (The Register, Fox Business, Los Angeles Times).
What the government showed the jury
Prosecutors presented evidence that Ding copied thousands of pages of confidential Google materials. Summaries differed: some reports cite “over 2,000 pages,” while investigators referenced roughly 14,000 pages across more than 1,000 unique files, with about 105 core documents identified.
Exhibits included internal logs and file-transfer records the government says were deliberately hidden. Testimony and exhibits reportedly showed Ding traveled to China between October 2022 and March 2023 for investor meetings tied to Rongshu, used a colleague’s badge to create a U.S. presence while abroad, and signed a December 2023 affidavit denying retention of Google files — an affidavit prosecutors say concealed uploads (Help Net Security, The Register, Fox Business).
Timeline of key events
- May 2022–April 2023: Prosecutors allege Ding copied and uploaded Google files (Help Net Security, Fox Business, Los Angeles Times).
- Oct 29, 2022–Mar 25, 2023: Travel to China and investor meetings tied to Rongshu (The Register).
- Dec 8, 2023: Signed Google affidavit denying retention of company data (The Register).
- Dec 14, 2023: Books one-way ticket to Beijing (The Register).
- Dec 26, 2023: Resigns from Google; Google suspends access after spotting a Zhisuan presentation (The Register).
- Jan 4–6, 2024: Google seizes devices; FBI executes search warrant (The Register).
- Mar 5, 2024; Feb 4, 2025; Sept 2025: Indictment and superseding indictments filed prior to trial (Fox Business, The Register).
- Jan 2026: 11-day trial ends in conviction (Fox Business, Los Angeles Times).
Potential penalties and immediate court action
Ding faces maximum statutory penalties of up to 10 years per trade-secret count and up to 15 years per economic-espionage count, plus potential large fines under federal sentencing rules — a theoretical aggregate that exceeds 170 years, though actual sentencing will follow guidelines and precedent (Help Net Security, The Register, Fox Business).
Judge Vince Chhabria released Ding pending sentencing, finding he was not a flight risk, and set a status conference for February 3, 2026 (DOJ, Los Angeles Times).
Reactions from officials and companies
Google’s Lee-Anne Mulholland, VP of Regulatory Affairs, said the company was “grateful to the jury” and warned that stealing trade secrets has serious consequences (The Register, Los Angeles Times).
U.S. Attorney Craig H. Missakian said the jury sent a “clear message” that the United States will protect its intellectual capital. FBI officials warned such theft threatens U.S. competitiveness in technology, and DOJ described the verdict as the first economic-espionage conviction tied directly to AI-related trade secrets (DOJ, Fox Business).
Defense attorney Grant Fondo said he respected the verdict but was disappointed, arguing there was no proof the data was sold or used and criticizing Google’s internal protections (Fox Business).
Economic impact
This case underscores a growing global worry: theft of AI trade secrets can accelerate technology transfer and alter competitive balances. Cloud providers, chipmakers and AI startups may face higher security and compliance costs, while European firms relying on outsourced infrastructure may reassess partnerships.
Companies may increase investments in data-loss prevention, audits and stricter supplier controls to protect sensitive hardware and software design details (DOJ, Los Angeles Times).
Political consequences
Governments may cite the verdict in debates on industrial policy, export controls and national security. European regulators and lawmakers already weighing AI rules could use this case to justify tighter oversight of dual-use technologies, talent recruitment programs, and cross-border collaborations (DOJ).
Social and legal effects
Public trust in cloud and AI platforms may be affected as users and businesses demand clearer assurances about monitoring and protecting sensitive technical work. The legal precedent of an economic-espionage conviction tied to AI data could spur similar prosecutions internationally and influence corporate compliance programs.
Cultural relevance and practical applications
The case will likely drive cultural shifts in handling internal access and remote work for staff with high-level system privileges. Practical responses may include stricter access controls, more frequent audits, limited remote access for critical infrastructure roles, and clearer whistleblower channels to prevent leakage of technical know‑how (The Register, Los Angeles Times).
Implications for AI policy
This conviction may accelerate policy moves to balance openness in research with stronger protections for sensitive infrastructure. Debates over the EU AI Act, export controls on advanced chips, and talent recruitment programs could intensify as governments seek to protect strategic capabilities while preserving innovation (DOJ).
