AI agents are still logging in as humans

Okta’s Enterprise AI Index analyzes how organizations increasingly run many AI platforms in parallel, creating identity and access risk as credentials, tokens, and agent workflows proliferate. The report documents growth of AI-native vendors and agentic features, warns that agents often inherit human logins (breaking auditability), and describes a larger unsanctioned “shadow AI” layer; it recommends least-privilege access, lifecycle management for non-human identities, and bringing popular AI tools into SSO and governance to reduce orphaned tokens and over-permissioned apps.