Human vs. AI Identity: Why AI Agents Are Breaking Identity

**Executive summary:** This report argues that autonomous AI agents fundamentally violate assumptions underlying traditional human and workload identity models—producing unpredictable access patterns, delegation/dual-identity complexity, and fragmented audit trails—and recommends a new, agent-aware identity approach that combines cryptographic user-agent binding, just-in-time/runtime policy evaluation with operation-scoped credentials, and agent-aware auditing that captures delegation chains and reasoning context to enable accountability.