China Investigates DeepSeek and Moonshot for Alleged Data Leaks to Anthropic

China Investigates AI Firms for Alleged Data Breaches

China's Cyberspace Administration (CAC) has launched an investigation into two prominent AI companies, DeepSeek and Moonshot, following allegations of unauthorized data transfers to Anthropic. This investigation is particularly significant as it coincides with both companies facing critical timelines.

According to a recent report from Anthropic, Moonshot allegedly funneled over 23 million exchanges to its AI model, Claude, through 5,380 fake accounts, while DeepSeek reportedly accounted for 12.1 million exchanges in a mere two-week span in July. These claims have prompted the CAC to question whether sensitive data from police, military, or state-affiliated entities was sent to servers in the United States.

The investigation was sparked by Anthropic's fourth threat intelligence report, released on September 10, which accused seven Chinese labs, including DeepSeek and Moonshot, of "illicit distillation"—a term used in AI to describe the practice of creating smaller models based on the outputs of larger models. While this method can be legitimate, Anthropic contends that it should not be executed through deceptive practices.

The CAC initially summoned all seven companies mentioned in the report for questioning but has since narrowed its focus to DeepSeek and Moonshot, with no penalties announced as of yet. This scrutiny comes at a crucial time for both firms; DeepSeek is preparing to brief the UN Security Council on AI risks this week, while Moonshot is gearing up for a $3 billion IPO in Hong Kong, aiming for a valuation of $50 billion. Any findings from the investigation will need to be disclosed in Moonshot's IPO prospectus, adding to the pressure on the company.