Okta enhances AI agent security with new platform features and Blueprint Alliance

As outlined in Silicon Angle, Okta Inc. has introduced significant enhancements to its Okta for AI Agents platform, including runtime enforcement and a broader kill switch, during its Oktane conference in Las Vegas. The company also announced its participation in the formation of the Blueprint Alliance, a consortium of 12 vendors aiming to establish an open, multivendor reference architecture for AI agent security.

Okta's new features, Agent Gateway and Shadow AI Agent Discovery for Endpoints, address the growing risks associated with AI agents gaining excessive access to sensitive systems. Agent Gateway acts as an intermediary, enforcing policies and logging interactions in real-time, while the planned kill switch will revoke active tokens and terminate sessions upon deactivation. Shadow AI Agent Discovery for Endpoints will identify unmanaged agents on employee devices. These additions aim to mitigate the dangers of powerful AI agents, especially when employees leave an organization, leaving ungoverned agents behind.

The Blueprint Alliance, which includes major players like AWS, CrowdStrike, and Google Cloud, is working to evolve Okta's initial agent security framework into an open standard. Gartner predicts a surge in AI agents, with enterprises expected to have over 150,000 agents in use by 2028, while a significant majority feel unprepared for AI agent governance. The alliance's framework will incorporate response strategies for compromised agents, emphasizing treating each agent as a distinct identity with scoped access.

Source: Silicon Angle