China Accused of Stealing AI Advances, U.S. Warned

The New York Post published an opinion piece on May 12, 2026, arguing that China's AI progress relies heavily on theft and smuggling. The NYPost reports that the White House accused Beijing of "industrial-scale" theft of know-how from American AI labs, and that U.S. prosecutors dismantled an international smuggling ring that allegedly funneled advanced chips to China in violation of export controls. The article also states that Chinese authorities blocked a $2 billion takeover by Meta of a Chinese AI start-up called Manus, and that Manus founders were prevented from leaving the country, according to the NYPost.

The NYPost describes the accused method as model "distillation," which the article defines as training smaller models on the outputs of larger, proprietary systems. The NYPost reports a leaked internal memo from Michael Kratsios, director of the White House Office of Science and Technology Policy, that says, "The US government has information indicating that foreign entities, principally based in China, are engaged in deliberate, industrial-scale campaigns to distill US frontier AI systems." The article further reports that, according to the NYPost, the U.S. AI company Anthropic detected 24,000 fraudulent accounts that produced more than 16 million exchanges with its systems, a figure the NYPost attributes to reporting on the incident.

Industry observers: Model extraction and distillation are well-recognized technical risks in the research community, and supply-chain interdiction of advanced semiconductors has been a recurring enforcement focus. Companies and governments frequently cite both online data-exfiltration techniques and hardware diversion as complementary channels that adversaries use to accelerate capability acquisition. For practitioners, the convergence of model-extraction vectors and chip-smuggling allegations reinforces the need to treat intellectual-property protection, API abuse monitoring, and supply-chain controls as overlapping security domains.

For analysts and practitioners, monitor formal U.S. government releases for corroborating evidence and named investigations; watch enforcement actions and indictments for technical details on extraction methods; and follow vendor disclosures about account-fraud and API-abuse mitigations. The NYPost opinion does not provide primary technical forensic reports, and the article frames these events through a political-opinion lens.