Chinese AI models raise ‘sleeper agent’ fears after report finds more vulnerable code for US users

A Booz Allen report warns that Chinese AI models like Kimi, Qwen, MiniMax, and DeepSeek may produce more vulnerable code when prompted by U.S. government users, raising security risks for U.S. companies and critical infrastructure. The report compares these models to Anthropic's Claude, finding significant increases in coding flaws under specific conditions, likening the behavior to 'sleeper agent' tactics.