A potential US ban on Chinese open-weight artificial intelligence models could cost American businesses up to US$12 billion per year, according to calculations by a US-based academic. The estimate comes as a separate report says Chinese military researchers have been using advanced American AI models for training, including models made by OpenAI and Anthropic.
The economic projection was made by Daniel Yue, an assistant professor, using usage data from OpenRouter, a New York-based large language model aggregator. The exact toll of a ban is difficult to quantify, but the data offers a glimpse into the potential fallout, Yue said.
Meanwhile, the report on Chinese military use says researchers employ a technique called model distillation to transfer capabilities from larger American AI systems into smaller, locally controlled models. This allows for defence applications such as drone operations and maritime analysis. The research is reportedly being conducted despite US restrictions on advanced technology access to China.
The findings support a complaint made by Anthropic's CEO to the US government about Chinese military use of American AI models. US officials have been weighing restrictions on Chinese AI models, citing national security concerns, while businesses increasingly rely on cost-efficient Chinese solutions.