United States of America: White House Office of Science and Technology Policy issued memorandum on adversarial distillation of US AI models (NSTM-4)

On 23 April 2026, the White House Office of Science and Technology Policy issued memorandum NSTM-4 to the heads of executive departments and agencies on adversarial distillation of US artificial intelligence (AI) models. The memorandum identifies what it characterises as industrial-scale distillation campaigns by foreign entities, principally based in China, that extract capabilities from US frontier AI models through the coordinated use of proxy accounts and jailbreaking techniques. The memorandum states that resulting models do not fully replicate the originals but allow foreign actors to release products performing comparably on select benchmarks at lower cost, and that such campaigns can be used to remove security protocols and mechanisms designed to ensure ideological neutrality and truth-seeking outputs. The memorandum sets out 4 commitments by the United States Government. Firstly, the United States Government would share information with US AI companies on the tactics and actors involved. Secondly, the United States Government would enable private sector coordination against such campaigns. Thirdly, the United States Government would work with industry to develop best practices for identifying, mitigating and remediating industrial-scale distillation. Fourthly, the United States Government would explore measures to hold foreign actors accountable. The memorandum took effect upon issuance.