Okta Expands Identity Controls for AI Agents

Okta announced three capabilities to manage AI agent identities and access: Agent Gateway (a mediation layer that verifies agents/users, evaluates policies, and brokers short-lived credentials), Agent-to-Agent Connections (controls and logs agent handoffs), and Resource Access Certifications for AI Agents (bringing agents into periodic access reviews). These features aim to keep credentials out of agents, improve traceability of actions to a human or agent identity, reduce exposure from prompt injection or model compromise, and align with emerging standards and NIST interest in agent identity and authorization.