U.S. Accuses China of Using Proxies to Extract American AI

According to Michael Kratsios, the White House science policy office says Chinese entities are using large-scale proxy networks and jailbreaking tactics to distil U.S. frontier AI systems, intensifying concerns over intellectual property and technology security.

Summary

Michael Kratsios, director of the White House Office of Science and Technology Policy, circulated a memo on April 23 accusing Chinese entities of carrying out deliberate, industrial-scale efforts to distil and steal U.S. frontier AI systems. According to the memo, the campaigns use tens of thousands of proxy accounts as well as jailbreaking techniques to extract American artificial intelligence. The allegation frames AI model extraction as both an intellectual property and national technology security issue. The source identifies the White House Office of Science and Technology Policy and Kratsios by name, but does not specify which Chinese entities were involved, which U.S. AI companies were targeted, what evidence was cited, or what specific U.S. response measures may follow.

Terms & Concepts
  • AI distillation: A technique for transferring behavior or outputs from one artificial intelligence model to another, which can raise concerns when used to replicate proprietary systems without authorization.
  • Proxies: Intermediary internet connections or accounts used to route and distribute online requests, potentially masking the origin or scale of activity.
  • Jailbreaking: Methods used to bypass artificial intelligence safety restrictions or guardrails in order to obtain otherwise blocked outputs or access.