End-to-end scalable red teaming and security testing for voice AI agents. Uses a proprietary, intentionally unaligned LLM that simulates manipulation tactics, exploits known flaws, and uncovers unknown vulnerabilities.