Okta and Google Cloud Partner to Secure AI Agents on Gemini Enterprise Platform

Okta has announced an expanded strategic collaboration with Google Cloud, introducing a unified identity architecture that brings enterprise-grade security and automated governance to artificial intelligence workflows. Through deep integrations with Google’s newly launched Gemini Enterprise Agent Platform and Chrome Enterprise, the partnership provides organisations with a platform-level mechanism to manage, audit, and restrict autonomous AI agents alongside traditional browser-based corporate environments.

The joint solution addresses a critical structural gap in modern cybersecurity programs. While enterprise generative computing tools expand quickly, organisational defence models have failed to keep pace with the unique risks of machine-speed automation:

• The Governance Gap: Approximately 92% of corporate executives report moderate or widespread use of AI agents within their local business infrastructure, yet only 34% of organisations apply the same rigid security and access controls to these automated tools as they do to human workers.

• Surging Token Exploits: Post-authentication identity attacks, such as browser session hijacking, have experienced a dramatic 127% year-over-year surge. Attackers increasingly bypass standard multi-factor authentication (MFA) by stealing active, browser-stored session tokens directly from user endpoints.

• Vendor Lock-In Risks: Platform flexibility remains an executive priority, with 62% of IT leaders identifying rigid single-vendor software ecosystems as a prominent strategic risk.

Extending Identity Governance to Autonomous Workforces

When an AI agent interacts with corporate databases, it inherits the access permissions of the account it represents. Unlike a human worker, an autonomous agent can perform thousands of data transactions across multiple distributed systems per minute without pausing to verify whether a structural command conforms to corporate data guidelines.

To bring order to this automated tier, the first phase of the collaboration wires Auth0 for AI Agents directly into the Agent Runtime layer of the Gemini Enterprise Agent Platform. This integration allows developers to embed access parameters directly into software applications without complex custom coding, leveraging several core runtime features:

• Token Vault Architecture: Securely manages downstream integrations by isolating, processing, and refreshing OAuth tokens inside a centralised vault, preventing agents from exposing sensitive keys across third-party environments.

• Human-in-the-Loop Checkpoints: Automatically pause an agent's execution path and trigger a mandatory human approval notification whenever an algorithm attempts a high-risk action, such as executing financial wires or altering system access permissions.

• Fine-Grained Authorisation (FGA): Restricts agents to the specific real-time permissions of the calling user, mitigating overprivileged processing behaviours and protecting private data boundaries.

• Auth for MCP: Supplies authentication and authorisation structures to any Model Context Protocol (MCP) server, allowing granular control over data tool ingestion.

To eliminate dangerous structural blind spots as organisations scale to tens of thousands of automated systems, Okta for AI Agents will soon deliver centralised visibility across the enterprise. The system continuously tracks and logs agents within an AI Agent Import & Registry directory, ensuring every agent is cryptographically bound to a verified human owner.

When external or internal agents attempt to interact with Google Cloud services, the platform routes requests through the Google Agent Gateway. The gateway acts as an air traffic control mechanism, delegating real-time authentication and authorisation checks back to Okta's central engine to ensure a single set of security policies governs both human and machine requests.

"Organizations shouldn’t have to choose between the AI and productivity tools their teams want and the security their business requires," stated Ely Kahn, Chief Product Officer at Okta. "Okta and Google are a natural fit because we pair Google’s leading product suite with an identity layer that can work across the entire modern, AI-powered work stack."

Hardening the Browser Layer Against Session Hijacking

Because modern enterprise workflows operate predominantly within web browsers, Okta and Chrome Enterprise have rolled out real-time threat response controls to secure the browser layer on both corporate and unmanaged endpoints:

A critical technical feature of this browser-hardening framework is native support for Device Bound Session Credentials (DBSC). Developed with Okta functioning as a core Google Cloud design partner, this open standard uses secure device-level hardware to lock an open web session to a specific physical endpoint through the Chrome browser.

While standard multi-factor authentication locks down the initial credential submission gate, DBSC ensures that even if a malicious browser extension successfully steals active post-authentication cookies, the exfiltrated session token is instantly invalidated if an attacker attempts to deploy it from an external device.

Vineet Bhan, Director and Global Head of Security and Identity ISV Partnerships at Google Cloud, emphasised the importance of a platform-level identity fabric:

"Securing the AI-powered enterprise requires a layer of identity security that operates seamlessly across the core platforms that power modern work. Together with Okta, we're extending that foundation across Google Cloud – so customers can confidently deploy AI agents in production, govern how they interact with critical systems, and maintain strong protection across the browser."

Read More:

The missing layer in enterprise AI: Why data trust is becoming the new priority

BVM 2026 Targets Global Trade Opportunities with 5,000 Foreign Delegates

Partner Pulse: Celebal Technologies | Cloud Partner (India)

Partner Pulse: TO THE NEW | Cloud and Digital Transformation Partner (India)