China has responded firmly to a recent US intelligence advisory accusing several Chinese companies of using distillation techniques to extract technological capabilities from American AI models. The Chinese government dismissed these accusations as unfounded smear campaigns and reaffirmed its commitment to self-reliance in AI development.
- US advisory accuses six Chinese firms of AI capability distillation from American models.
- China calls the accusations unfounded and claims AI progress is due to technological self-reliance.
- The dispute highlights broader geopolitical and commercial tensions ahead of US-China AI talks.
What happened
The US National Security Agency, FBI, and Cybersecurity and Infrastructure Security Agency jointly issued an advisory accusing six Chinese companies of employing distillation techniques at an industrial scale to replicate capabilities from US frontier AI models. The advisory detailed specific American models targeted and described the alleged circumvention of access controls via bulk-purchased premium subscriptions shared across teams.
In response, Mao Ning, spokesperson for China’s foreign ministry, dismissed the claims as baseless and politically motivated smears. She emphasized that China’s AI advancements are grounded in high-level technological self-reliance rather than illicit appropriation. Notably, the Chinese rebuttal did not engage directly with the technical or company-specific details presented in the advisory.
Why it matters
This exchange underscores the growing friction in AI technology development and policy between the US and China, two of the world’s leading AI powers. The US position frames distillation-based capability acquisition as a potential security risk and unfair competitive practice, while China maintains its narrative of sovereign innovation, a key theme in its broader technology strategy.
The advisory's detailed public attribution is unusual, signaling heightened concern about the transparency and legality of AI training methods in the international arena. For industry and regulators outside these two countries, the dispute complicates procurement decisions and regulatory scrutiny, particularly in regions like Europe where Chinese and American AI models compete directly on price and performance.
What to watch next
US-China discussions on AI governance scheduled for later this month will be pivotal. These talks are expected to focus more on broader issues such as export controls and market access rather than the contentious specifics of training data and distillation techniques. The current advisory exchange sets the diplomatic and technical background for these negotiations.
Observers should also monitor potential regulatory responses in global markets that use these AI models, particularly in Europe where questions about procurement integrity may arise. Meanwhile, the enduring industry-wide challenge around data licensing and model training could prompt further legal and policy developments affecting AI innovation globally.