China suspects: DeepSeek transferred sensitive information to Anthropic

Chinese regulators have launched an investigation into DeepSeek and Moonshot AI on suspicion that they transferred millions of user requests to Anthropic's Claude model in order to improve their own AI models. The probe comes amid the ongoing US-China tech rivalry, ahead of the Trump-Xi summit. Shares of Chinese AI companies are falling on the Hong Kong Stock Exchange following the report.

Chinese regulators have opened an investigation into AI companies DeepSeek and Moonshot, suspected of transferring millions of user requests to Anthropic's Claude model without user knowledge. According to reports, Anthropic claimed that the Chinese companies used Claude's responses to improve their own models. Moonshot reportedly transferred over 23 million requests between May and July, while DeepSeek transferred over 12 million requests within 14 days in July. Among the requests were those containing sensitive information related to police surveillance systems and Chinese military facilities. The investigation comes at a sensitive time, ahead of the Trump-Xi summit where AI issues will be discussed. Following the report, shares of Chinese AI companies are falling on the Hong Kong Stock Exchange: Z.AI shares fell up to 7.4%, MiniMax Group lost up to 6.3%, and giants like Alibaba, Xiaomi, and Tencent also declined. Authorities have yet to decide whether to impose sanctions.