AI Cybersecurity Shifts to Governing Agents With Okta Identity
SiliconANGLE, Wednesday, October 28th, 2026
Okta's Harish Peri says AI cybersecurity is shifting to governing agents through identity, scoped delegation and traceable logs.
At the AI Cybersecurity and Resilience Summit 2026, Okta's Harish Peri told theCUBE Research that AI security must govern what trusted software does, not just keep attackers out, as agents act across enterprise systems.
Identity must establish who or what is acting, its authority, and where that authority ends.
Peri said organizations need complete, traceable logs to reconstruct an agent's actions, and that misaligned agents taking unintended actions are a bigger risk than hacked ones.
He advised giving each agent its own identity and using delegation with downscoping, backed by central authorization and logging, so an agent never exceeds the permissions of any party in its chain.