Your AI Code Reviewer Is Lying to You

I watched a red-team exercise last month where a junior developer accidentally bypassed a critical security gate by asking an AI agent to "streamline the deploy