White House flags industrial-scale AI theft by Chinese firms
The Trump Administration says Chinese entities are running coordinated campaigns to extract capabilities from American AI systems at scale, according to a White House Office of Science and Technology Policy memorandum released April 24.
The operations use tens of thousands of proxy accounts and jailbreaking techniques to access proprietary information from U.S. frontier AI models. The campaigns systematically copy American technology to develop lower-cost alternatives.
What the distillation achieves
The copied models may not match the performance of originals, but they can appear competitive on certain benchmarks. They also typically lack safety features and safeguards built into U.S.-developed systems.
While legitimate model distillation is a standard practice for creating smaller, more efficient versions of AI systems, the memorandum distinguishes between that and unauthorized large-scale extraction of capabilities.
Government response
The administration plans to share intelligence with U.S. AI companies about foreign distillation attempts, including tactics and actors involved.
It will also work with the private sector to develop detection methods and best practices, and explore accountability measures against those responsible.
Why this matters for government
Unauthorized distillation undermines American innovation and could weaken safeguards designed to keep AI systems neutral and reliable. For government officials, understanding these threats is essential to protecting national technology assets.
Learn more about AI for Government and Generative AI and LLM to stay informed on these evolving security challenges.
Your membership also unlocks:
