China Dismisses US Distillation Advisory as Unfounded Smears
▼ Summary
– China’s foreign ministry rejected US intelligence claims that Chinese developers extracted capabilities from American AI models at industrial scale.
– Spokesperson Mao Ning asserted that China’s AI progress stems from technological self-reliance rather than the alleged distillation practices cited by Washington.
– The US advisory specifically named six Chinese companies and described their use of distillation as a core method, which Beijing did not technically refute in its response.
– Analysts note the diplomatic rejection avoids addressing specific technical allegations, likely due to upcoming high-level discussions between Trump and Xi on AI governance.
– The dispute highlights commercial competition, as Chinese open-weight models undercut American ones on price, while US labs face litigation over training data usage.
China’s foreign ministry has firmly rejected a recent joint intelligence advisory from the United States, dismissing claims that Chinese developers are industrially extracting capabilities from American frontier AI models as unfounded smears. The US document, issued by the NSA, FBI, and CISA, alleged that six specific Chinese firms were using distillation as a core method to build their systems rather than merely supplementing them. In response, spokesperson Mao Ning emphasized that China’s advancements in artificial intelligence are the product of technological self-reliance and domestic strength, urging Washington to cease such baseless accusations.
The US advisory was notable for its granular specificity, identifying particular companies and attributing targeted American models to each. It also challenged the widely reported training cost figures associated with DeepSeek. However, Beijing’s rebuttal did not engage with these technical details. Instead, it offered a high-level diplomatic assertion of national principle, mirroring the stance taken when the White House first raised concerns earlier this year. This approach highlights a recurring pattern in such exchanges: a technical accusation meets a political denial, leaving the factual disputes unresolved. The lack of a point-by-point refutation from either the ministry or the named companies leaves the specific allegations unaddressed.
The timing of the advisory adds another layer of complexity. With President Trump and President Xi scheduled to discuss AI governance later this month, neither side gains from a detailed public dispute over training data precedents. Meanwhile, US Treasury Secretary Scott Bessent, who has previously threatened sanctions on Chinese AI entities, stated that China can never surpass the United States. Such remarks naturally provoke strong responses from Chinese officials, regardless of the underlying technical merits.
Washington’s position rests on the distinction that distillation involves circumventing access controls, which differs from scraping publicly available web data. Yet, this argument faces scrutiny given the ongoing litigation against US labs over unlicensed training materials. Beyond the legal debate, there is a significant commercial dimension. Chinese open-weight models have undercut American competitors on price for two years, a trend observed by European buyers. By framing this competitive pricing as the result of theft, the US advisory serves both intelligence and market objectives. For European entities using both Chinese and American models, the advisory raises procurement questions, though no regulator has yet acted on these implications. Ultimately, the exchange sets the stage for upcoming negotiations focused on export controls and market access rather than resolving the immediate controversy over training data.
(Source: The Next Web)
