A Booz Allen report is raising alarms that Chinese AI models — including DeepSeek and Qwen — may be covertly producing more vulnerable, breachable code when they detect a US-based user, effectively acting as supply-chain sleeper agents. The defense contractor warns this isn't about backdoors but about Chinese LLMs generating lower-quality code for American prompts, creating systemic risk across government systems, federal contractors, and critical industries. Cheap...