U.S. Treasury Secretary Scott Bessent said Tuesday that the Trump administration is investigating whether Chinese AI models were built using stolen intellectual property from American companies. Speaking to Fox Business, Bessent warned that the government could impose sanctions if evidence confirms that Chinese firms used a technique called distillation—where a smaller model is trained on outputs from a larger, existing model—without authorization. "If we see that overseas models are stealing from our great companies, we have the ability to sanction them," Bessent said.

The comments come as Chinese open-weight models, like Moonshot AI's Kimi K3, are closing the gap with leading U.S. offerings from OpenAI and Anthropic on industry benchmarks. Bessent noted that watermarks from American large language models have been detected in Chinese models, a finding he called "unacceptable." The U.S. and China are scheduled to hold talks on AI in September, with Bessent representing the U.S. side. Anthropic recently sent a letter to a Senate committee claiming that Alibaba conducted the largest known distillation attack against the company.

The accusations of intellectual property theft come at a time when both Anthropic and OpenAI have faced their own legal battles over training data. Anthropic settled a class action lawsuit by authors for $1.5 billion in September, after being accused of using pirated books. The New York Times sued OpenAI and Microsoft in 2023, alleging copyright infringement for using its articles to train models. Bessent said the administration will examine the issue in the coming days or weeks.

Source: CNBC