Okta secures AI: Setting the Standard for the Secure Agentic Future

Okta CEO Todd McKinnon outlined a vision for the "secure agentic enterprise," arguing that while AI agents hold the potential for "genius" level innovation, they must be secured through strict identity management and standardization to avoid risk 1:51.

Key Takeaways:
• The industry faces a choice between locking down AI or letting it connect everything; Okta proposes "visibility and control" as the solution to enable both security and innovation 3:52.
• Okta formed the "Blueprint Alliance" with industry leaders like AWS and Salesforce to create a standard framework for securing AI agents, addressing the confusion caused by fragmented vendor solutions 8:08.
• A core component of this blueprint is "Agent SSO," which allows AI agents to authenticate to resources using enterprise identity policies rather than individual user consent prompts, streamlining access 21:59.
• Okta for AI Agents provides a unified registry where every agent has a unique, first-class identity with a named human owner, enabling granular governance and auditability 44:17.
• The platform includes an Agent Gateway that monitors runtime behavior, allowing security teams to detect and block data exfiltration attempts or unauthorized tool calls via a "kill switch" 50:12.
• Customer-facing agents, such as those used by the PGA of America, leverage these standards to provide seamless, personalized commerce experiences via protocols like UCP while maintaining secure authentication 53:34.

By standardizing how agents are identified and authorized across the technology stack, Okta aims to help enterprises safely deploy AI agents for both internal productivity and external customer engagement.

Sources:

• 1:51 McKinnon introduces the concept of "genius" AI and the industry's current focus on efficiency over innovation.
• 3:52 The speaker argues that visibility and control are the keys to balancing security with AI innovation.
• 8:08 Okta collaborates with AWS, Salesforce, and others to create a blueprint for securing AI agents.
• 21:59 Demonstration of Agent SSO allowing agents to authenticate to resources without manual user consent.
• 44:17 Explanation of Okta’s Identity Security Posture Management (ISPM) for discovering and registering agents.
• 50:12 Overview of the Agent Gateway’s role in monitoring runtime behavior and blocking unauthorized actions.
• 53:34 Example of PGA of America using Auth0 and UCP for secure, seamless customer-facing AI commerce.