Nvidia CEO Jensen Huang addressed the controversy surrounding AI model distillation during an interview on CNBC's Squawk Box. He characterized the practice of using outputs from existing models to train new ones as a standard form of market competition rather than theft.
Huang noted that companies frequently analyze and dismantle products to understand their underlying mechanisms. He acknowledged that while developers may prefer that others do not study their work, such competitive pressure ultimately drives industry improvements.
The CEO suggested that companies concerned about model distillation have the option to restrict access to their services. He stated that if a business does not want others utilizing their products in this manner, they should identify and block those specific clients.
The practice of model distillation has drawn significant scrutiny from US officials. Treasury Secretary Scott Bessent previously labeled it as theft in July and suggested potential sanctions against foreign firms that extract technical capabilities from American models.
In September, the US Cybersecurity and Infrastructure Security Agency accused Chinese AI firms of conducting large-scale knowledge distillation. Additionally, Anthropic alleged that Alibaba and DeepSeek engaged in unauthorized distillation involving their Qwen and DeepSeek models.
