China Rejects US Intelligence Advisory on AI as Unfounded Accusations and Smears
China’s foreign ministry has rejected a joint US intelligence advisory accusing Chinese developers of extracting capabilities from American frontier models at an industrial scale, calling on Washington to stop making unfounded accusations and smears.
Mao Ning, a ministry spokesperson, stated that China’s AI development is based on self-reliance, according to reports of her remarks. This response comes weeks before Trump and Xi are scheduled to discuss AI governance.
The US National Security Agency (NSA), Federal Bureau of Investigation (FBI), and Cybersecurity and Infrastructure Security Agency (CISA) named six Chinese companies, attributing specific American models to each and detailing how those companies allegedly built their AI through distillation. The document also challenged the training cost figure that initially brought DeepSeek into the spotlight.
However, China’s rejection does not address any of the individual company allegations, dispute the technical account of request routing, or respond to claims about bulk subscriptions shared across teams. It instead emphasizes self-reliance as a national principle, a common diplomatic response to such technical exchanges.
The advisory’s level of detail is unusual for a public intelligence report, making the discrepancy notable. Naming specific companies and attributing models invites a detailed rebuttal, which has not been provided by either the ministry or, publicly, by the companies themselves.
Timing plays a role, with Trump and Xi set to discuss AI governance later this month. Neither side appears to benefit from a detailed argument about data training practices in the lead-up to their meeting.
US Treasury Secretary Scott Bessent, who has previously threatened sanctions on Chinese AI, has stated that China can never surpass the US in this field. Such remarks prompt foreign ministry responses regardless of factual merits.
The challenge for Washington is that distillation itself is not inherently illegal, and the industry it protects has a history with others’ data. US labs face growing litigation over training material they did not license.
The advisory’s response acknowledges this distinction but remains narrow in its framing. There is also an unspoken commercial aspect neither government mentions. Chinese open-weight models have undercut American ones on price for two years, and European buyers have noticed. In Brussels, reporting on the competitiveness argument consistently reveals that the cheapest capable models are often not American.
An advisory labeling these price advantages as theft serves market interests alongside intelligence ones. For Europe, the practical concern is less about either country’s AI capabilities and more about the models in use within their own companies and public bodies.