BeyondTrust, a global leader in privilege-centric identity security, has unveiled AI Agent Security, a new module within its Pathfinder platform designed to govern and enforce what AI agents can do on enterprise endpoints before actions are executed.
The solution addresses growing concerns around the rapid adoption of AI coworkers and autonomous agents such as Microsoft Copilot, Claude Code, Cursor, and OpenAI Codex, which often inherit the same privileges as the users who launch them. According to BeyondTrust, these AI tools can access cloud, SaaS, and production environments at machine speed, creating significant security risks if left unchecked.
Built natively on the BeyondTrust Pathfinder platform, AI Agent Security provides organizations with a centralized framework to discover, manage, and control AI-driven activities. The platform enables security teams to identify approved and shadow AI tools across endpoints, define granular permissions, restrict access to external services and plugins, and prevent unauthorized or risky actions before they occur.
“We are not a privileged access company adding AI. We are the company that has long defined how to secure privileged action, and the most powerful actor on the endpoint is no longer human.” — Marc Maiffret, Chief Technology Officer, BeyondTrust
The module also introduces real-time runtime controls that can block activities such as credential exfiltration, deletion of production code, or database erasure across Windows, macOS, Linux, and containerized environments. Organizations benefit from a complete audit trail that records actions, identities, and permissions used by AI agents.
BeyondTrust said the solution builds on more than two decades of endpoint privilege management expertise and leverages research from its Phantom Labs security team. The company noted that non-human identities, including AI agents, are growing rapidly within enterprises, driving the need for stronger governance and privilege controls.
AI Agent Security is currently available in private beta and is expected to reach general availability on the Pathfinder platform in Fall 2026 as an add-on to Endpoint Privilege Management.
