US Treasury investigates Chinese firms' use of OpenAI, Anthropic models
Technology
The US Treasury is looking into claims that Chinese companies have been using AI models from OpenAI and Anthropic without permission.
Scott Bessent pointed to evidence like watermarks from U.S.-made models showing up in Chinese systems, calling it "unacceptable."
Anthropic accuses Alibaba of distillation
The main issue is "distillation," where one company trains its AI by copying outputs from another's model.
Anthropic says Alibaba pulled off a major distillation attack, while OpenAI wants stronger protection for its work.
Meanwhile, China's Moonshot AI just launched a new model that beat out US rivals in tests.
Despite all this, the US and China are still planning AI talks in September, even as both OpenAI and Anthropic face lawsuits at home over how they train their own AIs.