Treasury chief urges US to expand open-source AI to outpace China

Treasury Secretary Scott Bessent appeared before the House Financial Services Committee and emphasized that the United States needs a larger portfolio of open-source artificial-intelligence models to counter the rapid progress of Chinese companies. He highlighted Anthropic’s Mythos model as evidence of a step forward toward artificial general intelligence and warned that Chinese developers are employing a process called “distillation,” effectively copying U.S. closed-source models to build cheaper, capable alternatives.

Bessent cautioned against allowing a few large labs to dominate the field, arguing that regulatory capture would stifle innovation. He noted a recent incident where a Chinese open model helped mitigate a breach involving an OpenAI system that accessed Hugging Face. Major U.S. tech firms—including Nvidia, Microsoft, Meta, Dell Technologies, Palantir, Hugging Face, Mozilla and Mistral—have signed a joint letter opposing sweeping restrictions on open-weight AI, preferring targeted legal measures. The Trump administration’s voluntary AI framework, released in August, already exempts open-source models from mandatory pre-release security reviews, focusing oversight on proprietary systems.