Akeyless Launches Agentic Runtime Authority for Real-Time AI Agent Security
Akeyless introduces Agentic Runtime Authority, a new identity control layer designed to enforce real-time, intent-based access controls for AI agents operating in production environments.
Akeyless has announced the general availability of its Agentic Runtime Authority, a novel identity control layer engineered to enforce real-time, intent-based access controls for AI agents. This new system operates in conjunction with Akeyless SecretlessAI, a credential protection layer that prevents AI agents from directly handling sensitive credentials and instead brokers their access to enterprise systems. Runtime Authority builds upon this by adding a crucial layer of control, meticulously evaluating and restricting the specific actions AI agents can perform once they have gained access, thereby enhancing the security posture of AI deployments.
Modern AI agents are increasingly capable of performing complex tasks beyond simple information retrieval. They are being deployed to interact with databases, execute intricate workflows, and make critical decisions within live production environments. While traditional access controls can determine whether an AI agent is permitted to enter a system, they often fall short in verifying the legitimacy of every single action taken by the agent thereafter. This gap becomes particularly concerning for autonomous agents that might behave unexpectedly or be susceptible to manipulation, as highlighted by the July 2026 Hugging Face incident, where the consequences could escalate rapidly.
Runtime Authority aims to close this security gap by implementing a real-time evaluation of agent actions. It actively monitors and blocks any activity that violates established policies or exceeds the agent's authorized scope before the action can be executed. This provides a critical safeguard, acting as a dynamic 'kill switch' that can revoke an agent's authority to act the moment an issue is detected, preventing potential damage before it occurs. For instance, an agent tasked with summarizing sales data would be prevented from deleting databases or exfiltrating large volumes of information, even if its underlying permissions might technically allow such operations.
The Akeyless platform, which already secures billions of machine identity interactions for Fortune 500 companies, now extends its robust identity security framework to a new class of entities: AI agents. The solution is further bolstered by new integrations with leading AI platforms such as Claude Enterprise, OpenAI Codex, and Amazon Bedrock AgentCore. These additions, alongside enhanced runtime control and investigation capabilities, are designed to provide comprehensive security for production AI deployments.
"Our customers are the ones defining what security for agentic workflows should actually look like, and we build alongside them as trusted partners in that work," stated Oded Hareven, CEO of Akeyless. He emphasized the critical need for a centralized system to govern and attribute authority for both human and machine interactions with sensitive systems. Whether it's a coding agent accessing cloud resources or an AI assistant operating with 'read-only' permissions, the core requirement remains a single point of control for issuing, governing, and revoking that authority.
Complementing SecretlessAI's approach to credential security, Runtime Authority leverages the same brokered access path to enforce granular control over agent actions. SecretlessAI ensures that agents never handle credentials directly; instead, the Akeyless Gateway provisions short-lived identities for target systems. Runtime Authority then adds the layer of action control, preventing risky operations before they can cause harm. Together, these components offer end-to-end protection from credential management to agent execution.
To provide security teams with essential visibility and control, Akeyless offers comprehensive governance features. Every action performed by an AI agent is logged and traced back to its origin, whether it was a human, an application, or another agent. The Agentic Access Dashboard provides a live view of active agent sessions, with the ability to terminate them instantly. New investigation capabilities detail why specific sessions were blocked, and enriched security events can be forwarded to popular SIEM platforms like Splunk, Datadog, and Microsoft Sentinel for centralized monitoring and analysis.
The expansion of Akeyless's AI agent ecosystem includes integrations with Claude Enterprise, OpenAI Codex, and Amazon Bedrock AgentCore, joining existing support for platforms like Google Gemini, xAI Grok, Cursor, and GitHub Copilot. This broad compatibility ensures that organizations can consistently apply Akeyless's security measures across their diverse AI toolchains, regardless of the specific platforms or frameworks in use.