The US has raised alarm over what it described as “deliberate, industrial-scale” efforts by China-based entities to extract capabilities from American artificial intelligence systems, warning that such activities threaten both innovation and national security.In a memorandum to heads of executive departments and agencies, Michael J Kratsios, Assistant to the President for Science and Technology, said Washington has information indicating that foreign actors, “principally based in China”, are systematically distilling US frontier AI models.According to the Office of Science and Technology Policy, these campaigns involve the use of tens of thousands of proxy accounts to evade detection, along with “jailbreaking” techniques aimed at exposing proprietary model behaviour and safeguards.The memo described the effort as a coordinated attempt to extract capabilities from advanced American systems and replicate them at a fraction of the cost. While distilled models may not fully match original performance, they can mimic results on selected benchmarks and, in some cases, be deployed without critical safety and alignment mechanisms.“Such campaigns systematically exploit American expertise and innovation,” the memo noted, adding that some derived systems may deliberately strip away safeguards designed to ensure reliability, neutrality and responsible use.Washington drew a distinction between legitimate AI distillation — a widely used technique to create smaller, efficient models — and what it termed “unauthorised, industrial-scale” efforts aimed at undermining the US research and accessing proprietary information.Calling the trend “unacceptable”, the administration outlined a series of steps to counter the threat. These include sharing intelligence with US AI firms on tactics used by foreign actors, improving coordination between government and industry, and developing best practices to detect and mitigate such activities.The US also signalled that it would explore measures to hold those responsible accountable, though it did not specify potential penalties. “There is nothing innovative about systematically extracting and copying the innovations of American industry,” the memo said, warning that models built through such methods may ultimately lack integrity and reliability.The development comes as global competition in artificial intelligence intensifies, with Washington seeking to safeguard its technological edge while promoting what it calls a “free and fair” AI ecosystem spanning frontier systems, open-source frameworks and applied technologies.Reaffirming its broader strategy, the US said it would continue to support domestic industry, expand access to advanced AI tools, and strengthen market conditions that enable the responsible diffusion of emerging technologies worldwide.