Available on the BeyondTrust Pathfinder Platform, AI Agent Security provides real-time control over individual AI agent actions without restricting the same actions for human users.
BeyondTrust has announced the general availability of AI Agent Security on the BeyondTrust Pathfinder Platform, alongside new research from BeyondTrust Phantom Labs. The research examines AI agent activity and the connections these agents introduce across enterprise endpoints.
Phantom Labs Research Reveals Key AI Agent Security Findings
- Nearly 10 shell commands requested per prompt. AI coding agents requested approximately 1.28 million shell commands from 132,000 prompts, showing how a single prompt can trigger multiple downstream command requests. This highlights the need to monitor agent activity and enforce policies.
- 85% of observed coding-agent users had MCP server connections. Among users with coding agents configured, 85% had at least one MCP server configured, averaging 4.35 unique servers per connected user, with one user connected to 36. These connections can extend agents’ access across platforms and introduce security, privacy and compliance risks if not properly monitored.
- About half of agent skill installs arrive bundled in code repositories. In a mid-sized enterprise environment where 60% of AI-active devices ran a coding agent, a typical user with agent skills had approximately two dozen installed, and at least one in seven installed skills was used within four weeks. As skills can contain executable code and influence agent behaviour, organisations need visibility into installed skills and controls to identify unapproved or compromised ones
BeyondTrust says AI Agent Security distinguishes AI agent activity from human activity and evaluates agent actions against policies at runtime, allowing organisations to control what agents can do without changing the permissions employees need for their work.
“AI agents are increasingly acting on behalf of employees, often with the same access as the humans using them,” said Marc Maiffret, CTO of BeyondTrust. He said organisations need to control the privileges granted to AI agents separately from those assigned to their human users.
Join the Techitup Executive Network
Join the Middle East’s inner circle of C-Suite Tech Leaders. Exclusive access to closed-door regional roundtables, priority VIP seating at our annual awards, and strategic industry briefings.
Controlling AI Agent Activity
The platform provides visibility into AI agents operating across endpoints, including internally hosted models and AI tools that may be used without formal approval. It also records contextual information about agent activity, including the associated user, device, session, skills and Model Context Protocol (MCP) connections.
BeyondTrust says the product evaluates individual agent actions against policy before execution. Where additional privileges are required, it can invoke just-in-time access controls or human approval.
Availability
AI Agent Security is generally available on the BeyondTrust Pathfinder Platform for Windows and macOS endpoints.
Disclaimer: This article is based on an official announcement by BeyondTrust and its Phantom Labs research. It has been edited for length and readability. Product capabilities and availability are based on information provided by the company. Refer to the original research for detailed findings and methodology.
