The White House has formally accused China of engaging in what it describes as an 'industrial-scale' effort to distill and replicate American artificial intelligence models, raising new concerns about intellectual property theft and global AI competition. The Office of Science and Technology Policy (OSTP) released a policy memo outlining the findings and signaling a stronger approach to safeguarding U.S. AI innovations.
Allegations of Model Distillation
The memo highlights that Chinese entities are systematically extracting knowledge from leading U.S. AI models, a process known as 'distillation,' which involves training smaller models to mimic the behavior of larger, more advanced systems. This practice, according to the White House, undermines the competitive advantage of American companies and threatens national security interests. The ostensible scale of the operation has prompted officials to consider more robust accountability mechanisms.
Intelligence Sharing and Corporate Response
In response, the U.S. government has committed to sharing intelligence with domestic AI firms such as OpenAI, Anthropic, and Google. These companies have also been vocal about the issue. OpenAI, for instance, recently accused the Chinese startup DeepSeek of distilling its proprietary models, while Anthropic has identified DeepSeek, MiniMax, and Moonshot AI as creators of over 24,000 fraudulent accounts that generated more than 16 million AI-generated posts. These actions, according to Anthropic, are part of a broader strategy to artificially inflate their AI capabilities and mislead the public.
The White House's move underscores the growing importance of AI governance and the need for international cooperation to counter unauthorized model replication. As the U.S. pushes for stronger enforcement, it signals a broader shift in how it plans to protect its technological edge in the global AI race.
