The United States has signalled a significant escalation in its technological rivalry with China, with Treasury Secretary Scott Bessent warning that Chinese open-source artificial intelligence (AI) models could face sanctions over alleged intellectual property (IP) theft. Speaking on Tuesday, Secretary Bessent stated that the US would scrutinise these models for evidence of IP infringement, reserving the right to impose sanctions on Chinese AI companies if such theft is established.
This declaration marks a potential shift in strategy for Washington, which has previously focused on restricting China's access to advanced computing chips and tightening export controls. The new threat directly targets the AI models themselves, coming at a time when Chinese AI firms, such as Moonshot AI with its Kimi K3 model, are demonstrating increasing capabilities and market penetration. These advancements are seen by some as a growing challenge to the business models and fundraising abilities of leading American AI developers like OpenAI and Anthropic.
While the US administration has expressed support for the concept of open-source models, Secretary Bessent drew a clear line regarding IP theft. "This administration supports open source models, but what we do not support is IP theft," he remarked. "If we see, especially, that overseas models are stealing from our great companies, we have the ability to sanction them because of this theft." The comments, first reported by Bloomberg, underscore a deepening concern within the US government about maintaining its lead in the global AI race.
However, the definition of what constitutes IP theft in the context of AI model development remains a point of contention within the industry. Techniques like 'model distillation', where some capabilities of a larger model are transferred to a smaller, more efficient system, are widely used. Microsoft CEO Satya Nadella has previously criticised large labs for presuming distillation automatically equates to theft, arguing for fair use rights to train models on public data while questioning restrictive terms on distillation. Furthermore, some experts, including Hugging Face CEO Clem Delangue, suggest that distillation is a minor factor in China's AI progress, attributing their advancements more to strong research teams and a collaborative approach.
For the UK, this development could have several implications. British companies involved in AI development or those utilising Chinese AI technologies may need to assess their supply chains and partnerships. While the immediate focus is on US-China relations, any broader restrictions on AI model exchange could impact global collaboration and the availability of diverse AI tools. The UK government, through the Foreign Office, typically advises British nationals and businesses to monitor international sanctions and trade policies, and this new US stance could necessitate a review of existing and planned AI collaborations involving Chinese entities.