The White House has accused foreign entities, predominantly in China, of orchestrating “industrial-scale” theft of American artificial intelligence models through coordinated distillation attacks. These foreign actors are reportedly utilizing tens of thousands of proxy accounts and jailbreak techniques to extract proprietary capabilities from U.S. AI systems, with evidence indicating that these efforts could bypass security safeguards aimed at preserving the integrity of these technologies. In response, the Trump administration is directing federal agencies to collaborate with U.S. AI companies, such as OpenAI and Anthropic, to bolster defenses against such unauthorized extrications and ensure accountability for these foreign activities.

White House: The White House serves as the executive office of the President of the United States under President Donald Trump, coordinating federal policies on science, technology, and national security. Its Office of Science and Technology Policy recently issued a memorandum accusing foreign entities primarily based in China of conducting industrial-scale distillation campaigns to steal capabilities from American frontier AI models. The administration pledged collaboration with federal agencies and private industry to build defenses and pursue accountability for these adversarial activities.
Michael Kratsios: Michael Kratsios is the Assistant to the President for Science and Technology and the 13th Director of the White House Office of Science and Technology Policy. He advises on advancing U.S. innovation in AI and other technologies, drawing from prior roles including U.S. Chief Technology Officer. Kratsios authored the April 2026 memo ‘Adversarial Distillation of American AI Models,’ detailing evidence of coordinated foreign theft efforts mainly from China and calling for protective actions.

`json
{
“Policy Action”: “The Trump administration directs federal agencies to partner with the private sector to fortify AI model protections against unauthorized extraction by Chinese actors.”,
“Industry Response”: “U.S. AI companies, including OpenAI, Anthropic, and Google, share threat intelligence through the Frontier Model Forum to detect large-scale distillation attempts.”,
“Distillation Threat”: “Distillation attacks enable foreign entities to train smaller AI models that mimic the outputs of larger U.S. frontier systems, potentially bypassing security safeguards.”
}
`