Okta and Google Cloud are deepening their partnership to secure AI agents within both the cloud environment and Chrome.
Okta and Google have long been partners in cloud security, and both parties are further expanding their collaboration with new integrations. These bring Okta’s identity management platform to Google’s Gemini Agent Enterprise Platform. The integration is designed to ensure the secure deployment of AI agents within corporate cloud environments.
That is often where things go wrong. Okta previously shared a study showing that rules are less stringent for AI agents than for human employees. While 92 percent of companies reported using AI agents, only 34 percent apply security controls. While human employees must constantly log in, AI agents are often given free rein.
Integration into Gemini Enterprise
Okta will add a dual layer of identity security to Gemini Agent Enterprise. An initial integration, available immediately, implements OAuth authentication in AI agent workflows without the need to develop extra code. The Token Vault stores and assigns authentication tokens to AI agents so they can gain access on behalf of human users. Human approval is required for ‘high-risk’ tasks.
Later, Okta will also add its security platform for AI agents to Google Cloud for centralized visibility and control. In a central dashboard, AI agents can be linked to human employees. Requests and authentication are exchanged and approved between the Okta platform and Google Cloud’s Agent Gateway.
Chrome
Okta is also keeping an eye on the Chrome browser. Google is injecting more and more AI functionality into its browser, and in its future vision, AI agents will be able to browse the web freely. Okta’s role is to help ensure this happens in a controlled manner.
A new ‘enrollment module’ will be added to the Okta Integration Network platform, allowing policy rules to be added to managed Chrome profiles without requiring synchronization with the Google profile. Okta is also integrating with the Device Trust Connector in Chrome to block suspicious login attempts if the device’s antivirus fails. Active sessions are linked to a device so that stolen cookies cannot be used on another device to hijack a session.