Counsel-ready red-team audits for deployed voice AI. Statute-mapped adversarial calls scored on long-horizon dialogue, compliance, malicious intent, inter-agent comms, and language switching.