The Trump administration has accused China-backed actors of running

deliberate, industrial-scale campaigns
to replicate U.S. frontier AI models, according to a memo from Michael Kratsios, director of the White House Office of Science and Technology Policy. The memo, sent to federal agency heads, alleges that China-based entities are using proxy accounts to evade detection and extract proprietary information from American AI systems.

Escalating AI Rivalry

These distillation attacks involve querying proprietary models, such as Claude or Gemini, millions of times via APIs to build datasets that replicate the behavior of U.S. systems. Kratsios warned that such tactics enable foreign actors to release models that mimic U.S. capabilities at a fraction of the cost, while also undermining guardrails designed to ensure outputs remain ideologically neutral and truth-seeking.

The accusation comes as President Trump prepares for a high-stakes visit to Beijing next month, where he is expected to push for economic concessions and recalibrate aspects of the U.S.-China relationship. The warning also follows reports from OpenAI and Anthropic, which earlier this year identified China-based firms, including DeepSeek, Moonshot AI, and MiniMax, as being behind wide-scale distillation attacks on their models.

Broader Context

This is not the first time the U.S. has accused China of intellectual property theft. In 2024, the Justice Department indicted a former Google software engineer for allegedly stealing AI trade secrets and sharing them with Chinese companies. However, Kratsios noted that the reliability of these distilled models may degrade over time as detection methods improve.

The Trump administration plans to share intelligence with U.S. AI companies to help them develop defenses against such campaigns, Kratsios added. This move underscores the administration's focus on safeguarding American technological dominance in the face of growing global competition.