The Same Zero: Why Identical ASR Can Imply Different Guarantees in LLM-Agent Security

LLM-agent security has produced a dense landscape of defenses - prompt hardening, content filters, permission gates, sandboxes - yet no framework tells a deployer what a defense actually guarantees, or where that guarantee comes from. We apply Verification Autonomy Levels VAL - L0: LLM...