AI Agents: The Next Wave Identity Dark Matter - Powerful, Invisible, and Unmanaged

**Executive Summary:** This article warns that widespread enterprise adoption of MCP-enabled AI agents is creating 'identity dark matter'—invisible non-human identities (orphaned local accounts, stale service identities, long-lived tokens, API keys, and bypass auth paths) that agents will exploit for efficiency—leading to over-permissioned access, untracked usage, privilege drift, and regulatory blind spots; it outlines typical abuse patterns and prescribes five core principles (human sponsors, dynamic context-aware access, visibility and auditability, enterprise-scale governance, and IAM hygiene) to safely adopt agentic systems.