Chinese regulators are investigating DeepSeek and Moonshot AI after Anthropic reported that the firms had illicitly used its Claude model for training. Investigators are examining whether the companies sent sensitive user data—including police and military surveillance information—to Anthropic’s U.S.-based servers without users’ consent, potentially violating China’s data security laws.
Separately, Anthropic is asking shareholders to approve a governance change that would give its founders 50.1% of the voting power through a separate class of shares. The proposed structure resembles the dual-class voting setup at Peter Thiel’s Palantir.
The developments come amid broader debate over AI safety and security at Anthropic. Those concerns could make it harder for the company to attract and retain enterprise customers.