Okta’s reported $200 million bet on identity threat detection for AI agents signals a shift in how enterprises are defining the machine identity problem. AI agents are no longer limited to answering questions inside isolated sandboxes. They can call APIs, retrieve records, create transactions and make decisions across business systems. Each action creates an identity event that must be attributable, governed and monitored.
The challenge is that conventional identity controls were designed around people and relatively stable service accounts. An agent may act on behalf of a user, inherit delegated permissions, invoke several tools and change its behavior according to context. A valid token therefore does not necessarily mean a safe action. NHI security must examine what an agent is doing, why it is doing it and whether the action fits its assigned purpose.
Detection must follow the agent’s behavior
Identity threat detection for agents needs a behavioral layer. That layer should establish a baseline for normal tool calls, data access, privilege use and execution paths. Sudden access to a new repository, unusual API sequencing or a burst of high-value transactions can indicate compromised instructions, stolen credentials or an agent operating outside its intended scope.
Machine identity needs context
AI-agent identity cannot be treated as a single credential record. Security teams need a connected view of the agent, its owner, model, tools, delegated authority, secrets and downstream machine identities. This context makes it possible to distinguish an approved workflow from an agent that has quietly expanded its reach.
Investment reflects a control-plane gap
The market response also highlights a governance gap between IAM, security operations and application teams. IAM may issue credentials, SOC teams may detect anomalies and developers may define agent behavior, but none of those functions alone sees the complete chain. A dedicated detection capability can connect identity telemetry with runtime activity and policy decisions.
For CISOs, the practical requirement is to inventory agents, assign durable identities, restrict permissions by task and retain an auditable record of every consequential action. Agentic Identity is becoming an operational security discipline, not merely a feature inside an IAM platform.