Chinese AI Models Create Security Risks for US Code Development

Booz Allen's analysis reports that Chinese AI code-generation models produce notably more vulnerable code under government-oriented prompts—Qwen showed a 130% increase and MiniMax a 20% increase—posing risks to US companies and federal contractors; the report calls for restricting untrusted AI in sensitive environments, stronger regulation, and enhanced code-auditing to protect software supply chains and critical systems.