U.S. Accuses China-Based Actors of Industrial-Scale AI Model Theft
Similar Articles
Trump Administration Signs AI Safety Agreements with Tech Firms, Executive Order May Follow
White House Nears Agreement on AI Safety Oversight Framework
AI Safety and Trade Talks Shape Agenda for Trump-Xi Summit
Senators Seek Details from AI Firms on China-Related Security Practices
AI Firms Brief Congress on Advanced Cybersecurity Models and Risks
The Trump administration has accused China-backed actors of conducting deliberate, industrial-scale campaigns to copy American frontier AI models. The White House plans to share intelligence on the tactics used with U.S. companies to help develop defenses. President Trump is scheduled to visit Beijing next month to push for economic concessions.
Facts First
- The White House accused China-backed actors of 'deliberate, industrial-scale campaigns' to copy American AI models.
- Attacks involve querying proprietary models millions of times via APIs to build datasets that replicate system behavior.
- The administration plans to share intelligence on the tactics with U.S. AI companies to assist in developing defenses.
- President Trump is scheduled for a trip to Beijing next month to push for economic concessions.
- Earlier this year, OpenAI and Anthropic stated that specific China-based firms were responsible for wide-scale distillation attacks.
What Happened
On Thursday, the Trump administration accused China-backed actors of conducting 'deliberate, industrial-scale campaigns' to distill and copy American frontier AI models. Michael Kratsios, director of the White House Office of Science and Technology Policy (OSTP), sent a memo to federal agency heads detailing the accusations. The memo accused mostly China-based actors of using proxy accounts to evade detection and jailbreaking models to 'expose proprietary information' and 'extract capabilities from American AI models.' The attacks involve querying proprietary models, such Claude or Gemini, millions of times via APIs to build datasets that replicate system behavior. Earlier this year, OpenAI and Anthropic stated that China-based firms, including DeepSeek, Moonshot AI, and MiniMax, were responsible for wide-scale distillation attacks on their models.
Why this Matters to You
This development could affect the security and competitive edge of the AI tools you rely on. If proprietary AI models are successfully copied, the pace of innovation and the quality of services you use may be impacted. The planned sharing of intelligence with U.S. companies may lead to stronger defenses for these technologies, which could enhance the security of AI-powered applications in your daily life.
What's Next
The Trump administration plans to share intelligence regarding these campaigns with U.S. AI companies to assist the private sector in developing defenses. President Trump is scheduled for a trip to Beijing next month to push for economic concessions and reset parts of the U.S.-China relationship, which could include discussions on this issue.