Okta: AI agents become a new source of insider security risk

theCUBE Research analyst Krista Case said at Okta's Oktane event that AI agents create a new class of insider risk for enterprises. One financial asset manager discovered about 13,000 agents in its environment but deemed only 1,000 valid. Okta's approach covers discovery, authorization and runtime controls, though a kill switch cannot undo actions an agent already took.

• A financial firm found 13,000 AI agents and considered only 1,000 valid
• A kill switch revokes authority but doesn't undo changes an agent already made
• Agents shouldn't automatically inherit all access and permissions of their users
• Okta's Blueprint Alliance aims for a shared architecture for agent control