Anthropic says it shut down multiple state-sponsored surveillance operations that used its AI models. In a report released Thursday, the San Francisco-based company says the misuse was detected and disrupted between January and July, and that it came from China, Iran and West Africa.
Anthropic warns that governments and state-aligned actors increasingly use AI to target ethnic minorities and dissidents, including diaspora communities. It says some campaigns aimed at groups such as pro-democracy figures in Hong Kong and Tibetan and Falun Gong communities, as well as Iranian minority groups and opponents of the Iranian regime abroad.
Different outlets focus on specific case details mentioned by Anthropic. Several report that Iranian actors used a method to identify people via social media accounts, while a contractor working with Malian national security authorities used Claude to help develop underlying intelligence-gathering software. Multiple outlets also repeat Anthropic’s allegations that China’s Moonshot and Deepseek rerouted user requests to Anthropic through fraudulent accounts while “distilling” responses to train other models. In response, China’s foreign ministry says it opposes attacks and smears while supporting AI for good.
