US and China Plan AI Talks Amid IP Theft Allegations

The United States and China are planning bilateral talks in mid-September to address safety risks and the regulation of frontier artificial intelligence models. These discussions, a result of a May summit between President Donald Trump and President Xi Jinping, are intended to occur before the two leaders meet in Washington on September 24. U.S. Treasury Secretary Scott Bessent is expected to lead the American delegation.

The talks come amid escalating tensions over intellectual property. Treasury Secretary Bessent announced an investigation into whether Chinese models were created by stealing U.S. technology through distillation, a process of training smaller models using outputs from more powerful ones. Bessent stated that American watermarks have been found on Chinese models, describing the situation as unacceptable and threatening sanctions against offending entities.

These government actions follow reports from Anthropic, which accused Alibaba of executing a massive distillation attack using 25,000 fraudulent accounts. Other Chinese firms, including Moonshot AI, DeepSeek, and MiniMax, have also been implicated in similar campaigns. Specifically, White House advisor Michael Kratsios accused Moonshot AI of stealing from Anthropic's Fable model.

While the U.S. seeks cooperation on monitoring AI-directed cyberattacks and preventing the proliferation of powerful models to non-state actors, China is seeking technical engagement regarding the U.S.-developed Mythos model. In response to U.S. pressure, Beijing is reportedly considering curbing overseas access to its own top AI models to prevent technology leaks.