Bessent: US May Sanction China Over AI Model ‘Theft’

U.S. Treasury Secretary Scott Bessent has signaled a deepening concern over the competitive landscape in artificial intelligence, vowing to scrutinize whether Chinese AI models are being derived from American intellectual property. The administration’s stance is clear: “IP theft is not supported,” Bessent stated recently, underscoring a growing tension in the global AI race.

This move comes as Chinese open-weight models are rapidly gaining traction, posing a significant challenge to the established dominance of American AI giants like OpenAI and Anthropic. The emergence of sophisticated models, such as Moonshot AI’s Kimi K3, which has demonstrated superior performance on several industry benchmarks this month, is sparking anxieties among tech leaders and policymakers alike regarding the sustainability of U.S. leadership in this transformative field.

“If we see, especially that overseas models are stealing from our great companies, we have the ability to sanction them because of this theft,” Bessent asserted, highlighting the potential for retaliatory measures against entities found to be infringing on U.S. intellectual property. The mechanism he alluded to, known as “distillation,” is a technical process where a smaller, more accessible AI model is trained using the outputs of a larger, more powerful one. This method, while a common practice in AI development, raises red flags when applied without proper licensing or attribution.

The Treasury’s focus on distillation is particularly timely. Last month, Anthropic reportedly sent a letter to the U.S. Senate Committee on Banking, Housing, and Urban Affairs, alleging that Chinese tech conglomerate Alibaba engaged in what was described as the “largest known distillation attack” against the company. This accusation points to a potentially sophisticated and large-scale effort to replicate or leverage advanced AI capabilities.

These escalating concerns are set against the backdrop of planned U.S.-China dialogues on AI, slated for September. Secretary Bessent is expected to represent the United States in these crucial discussions, signaling the high-level importance Washington places on navigating the complexities of AI development and regulation with Beijing.

“We are finding watermarks of our U.S. large language models on many of the Chinese models, and that’s unacceptable,” Bessent stated, indicating that his department will be actively investigating these findings in the coming days and weeks. The presence of these “watermarks” suggests a direct, traceable link to U.S.-developed models, which the administration views as a clear violation of intellectual property rights.

However, the narrative of intellectual property disputes in AI is not one-sided. Both Anthropic and OpenAI have themselves faced allegations of copyright infringement. Anthropic reached a significant settlement in September, agreeing to pay $1.5 billion to a group of authors who claimed their copyrighted works were unlawfully downloaded and used for AI training. Similarly, The New York Times initiated legal action against OpenAI and Microsoft in 2023, alleging that the companies infringed on its intellectual property rights by incorporating its published content into their AI models without permission.

The ensuing legal battles and governmental scrutiny underscore the immense value and sensitivity surrounding AI development. As nations and corporations pour billions into AI research and deployment, the protection of intellectual property and the establishment of fair competition standards are becoming paramount. The U.S. Treasury’s proactive stance suggests a strategic effort to safeguard its technological lead and ensure a level playing field in the global AI economy, potentially setting new precedents for international cooperation and regulation in this rapidly evolving sector.