US Agencies Allege Chinese AI Firms Engaged in 'Industrial-Scale' Model Distillation
US security agencies accuse Chinese AI firms of industrial-scale distillation, a cost-cutting method that trains smaller models using outputs from larger ones.
American cyber and law enforcement officials have formally accused Chinese artificial intelligence companies of engaging in what they describe as "aggressive, industrial-scale distillation activities." The charge was made in a joint statement issued by three US security agencies on Tuesday.
Distillation is a technique used in AI development where a smaller, less expensive model is trained using the outputs generated by a larger, more costly system. This approach allows developers to significantly reduce the financial burden associated with training new AI tools.
The accusation suggests that Chinese firms have been systematically leveraging this process to accelerate their own AI development while avoiding the high costs typically incurred in building foundational models. The statement from the US agencies frames these activities as a coordinated and large-scale effort rather than isolated instances of research.
The announcement adds to the growing friction between Washington and Beijing over advanced technology, particularly in the realm of artificial intelligence, where both nations are vying for strategic dominance.