Taming Agentic AI Risks Requires Securing Non-Human Identities

**Executive summary:** The article outlines how non-human identities (NHIs)—including service accounts, API keys, tokens, and AI agents—have proliferated and now dominate corporate attack surfaces, creating blind spots for ownership, lifecycle management, and monitoring; industry surveys and reports show increasing ratios of machine identities and frequent compromises or leaked secrets. It emphasizes that the problem is organizational as much as technical, suggests leveraging and evolving existing standards (OAuth/OIDC), improved secrets management, auditable ephemeral credentials, and coordinated industry practices to regain visibility and control.