Okta Unveils AI Agent Runtime Enforcement and Forms Blueprint Alliance with AWS, CrowdStrike

Okta Inc. introduced new runtime enforcement features for its Okta for AI Agents platform at its Oktane conference in Las Vegas, alongside the formation of the Blueprint Alliance—a coalition of 11 vendors aimed at creating an open standard for AI agent security. The updates include an Agent Gateway for real-time policy enforcement and a kill switch to immediately deactivate compromised agents. Okta also set an ambitious target to save 1 million employee hours this year using AI and automation.

Runtime enforcement and kill switch for AI agents

Okta Inc. unveiled new runtime enforcement capabilities for its Okta for AI Agents platform during its Oktane conference in Las Vegas, according to SiliconANGLE. The centerpiece of this update is the Agent Gateway, which operates in the execution path between an AI agent and the tools it accesses. The gateway enforces security policies in real time and logs each interaction for auditing purposes.

The company also introduced a kill switch feature, allowing administrators to deactivate AI agents instantly. When triggered, the kill switch revokes all active tokens and terminates ongoing sessions, mitigating potential security risks. This extends Okta’s existing System Log, which captures agent events for post-incident review in security information and event management (SIEM) systems.

Expanding visibility and control

Okta’s platform now detects both managed and unmanaged AI agents. Known agents can be registered directly or imported from platforms like Amazon Bedrock and Salesforce Agentforce. A new feature, Shadow AI Agent Discovery for Endpoints, identifies unmanaged agents running on employee devices, addressing risks associated with unsanctioned AI tools.

To further tighten security, Okta introduced Agent SSO, which extends Cross App Access—a feature launched in June 2025—to all single sign-on customers. This enables enterprises to replace non-expiring API keys with short-lived tokens tied to specific identities, reducing the risk of credential misuse.

Other enhancements include Agent-to-Agent Connections, which governs interactions between AI agents and logs each handoff for auditing, and Resource Access Certifications, which reviews agent permissions over time to prevent excessive access.

Blueprint Alliance: A multivendor approach to AI security

Okta announced the formation of the Blueprint Alliance, a coalition of 11 vendors, including AWS, CrowdStrike, Google Cloud, Databricks, and Salesforce. The alliance aims to create an open, multivendor reference architecture for AI agent security, addressing gaps in governance and response protocols.

Members have agreed on core principles, such as treating AI agents as first-class identities and scoping their access to specific tasks. The alliance also outlines a standardized response framework for compromised agents, including token revocation, session termination, and network quarantine, followed by a staged, auditable restoration process. Industry observers noted that the alliance’s success may depend on its ability to translate these principles into interoperable tools across vendors.

Nonprofit World Central Kitchen and appliance manufacturer GE Appliances are serving as strategic advisers to the alliance, providing enterprise and operational perspectives.

AI governance and enterprise adoption

The Blueprint Alliance cited Gartner research predicting that the average global Fortune 500 enterprise will deploy over 150,000 AI agents by 2028. Despite this growth, only 13% of organizations believe they have the right governance frameworks in place to manage these agents securely.

Okta’s own Dex AI agent exemplifies how enterprises are integrating AI into workflows. Dex serves as a unified interface for employees, routing tasks to specialized sub-agents while enforcing identity controls via Universal Directory and Okta for AI Agents. Jenna Cline, Okta’s Senior Vice President of Technology, Data, and Intelligence, highlighted how Dex acts as a proving ground for emerging features before they are released to customers.

Productivity gains and roadmap

Okta reported surpassing its 2025 goal of saving 250,000 employee hours through AI and automation. The company has since raised its 2026 target to 1 million hours, reflecting confidence in its ability to scale productivity gains. These efforts align with Okta’s broader strategy to prioritize technology investments that streamline operations and reduce manual workloads.

Several new features, including Agent SSO and Resource Access Certifications, are now generally available. Others, like enhanced runtime enforcement capabilities, are slated for release in Q3 and Q4 2026.

Collaborative security in enterprise AI

The rise of AI agents in enterprise environments has introduced new security challenges, particularly around identity management and access control. Traditional governance frameworks, designed for human users, often fall short when applied to autonomous or semi-autonomous agents operating at scale.

Industry analysts have emphasized the need for open standards and multivendor collaboration to address these gaps. The Blueprint Alliance reflects a growing trend of cross-industry partnerships aimed at creating shared security frameworks for emerging technologies.

What this means

LazyFounders analysis — our interpretation, not reported fact.

Okta’s announcements at Oktane signal a proactive response to the rapid proliferation of AI agents in enterprise environments. The introduction of runtime enforcement and the kill switch addresses a critical gap in AI governance: the ability to monitor and control agents as they execute tasks, rather than relying solely on post-incident logs. This shift from reactive to real-time security could set a new standard for how enterprises manage AI-driven workflows.

The formation of the Blueprint Alliance is equally significant. AI security is not a problem any single vendor can solve in isolation, and Okta’s decision to collaborate with competitors and complementary platforms like AWS, CrowdStrike, and Google Cloud reflects a maturing understanding of the challenges ahead. If successful, the alliance could accelerate the adoption of AI agents by providing enterprises with a clearer path to secure, interoperable deployments. However, the real test will be execution—translating high-level principles into tools that work seamlessly across diverse ecosystems.

Okta’s use of its Dex AI agent as an internal testbed is a smart move. It aligns product development with real-world use cases, ensuring that features like runtime enforcement and identity controls are battle-tested before reaching customers. This approach also underscores a broader trend: enterprises are no longer just building AI tools; they’re using them internally to validate their own products. For Okta’s customers, this means the platform may evolve in ways that are directly informed by the pain points of large-scale AI deployment.

The productivity targets—saving 1 million employee hours this year—are ambitious but achievable if AI tools can reliably automate repetitive tasks. For founders and operators, Okta’s focus on time savings highlights a key value proposition of AI agents: they’re not just about innovation for its own sake, but about tangible operational efficiencies. The challenge for enterprises will be integrating these agents into existing workflows without disrupting productivity during the transition.

Key takeaways

• Okta’s new AI agent runtime enforcement features aim to address security and governance challenges in enterprise AI adoption.
• The Blueprint Alliance reflects a collaborative approach to standardizing AI agent security across multiple vendors.
• Okta’s Dex AI agent serves as an internal testbed for identity controls, aligning product development with real-world use cases.
• AI-driven productivity gains are a core focus for Okta, with ambitious targets for employee time savings.

FAQ

What is Okta’s Agent Gateway?

The Agent Gateway is a runtime enforcement tool that sits between an AI agent and the tools it accesses. It enforces security policies in real time and logs every interaction, allowing enterprises to monitor and control agent behavior as tasks are executed.

What is the Blueprint Alliance?

The Blueprint Alliance is a coalition of 11 technology vendors, including AWS, CrowdStrike, and Google Cloud, formed to create an open, multivendor reference architecture for AI agent security. The alliance aims to standardize governance, response protocols, and identity management for AI agents.

How does Okta’s kill switch work?

The kill switch allows administrators to deactivate an AI agent instantly. When triggered, it revokes all active tokens and terminates ongoing sessions, preventing further access or actions by the agent.

What is the significance of the Gartner prediction about AI agents?

Gartner predicts that by 2028, the average global Fortune 500 enterprise will have over 150,000 AI agents in use. However, only 13% of organizations believe they have the right governance frameworks in place, highlighting the urgency for standardized security and management tools.

How is Okta using AI internally?

Okta uses its Dex AI agent as a unified interface for employees, routing tasks to specialized sub-agents. Dex also serves as a testbed for evaluating identity control technologies, such as Universal Directory and Okta for AI Agents, before they are released to customers.

Related on LazyFounders

• Amazon Web Services Inc.
• Proofpoint and Anthropic Outline Shift to Governing AI Agents in Enterprises
• PGA of America Overhauls Identity Security to Manage AI Sprawl
• More technology news