Bitdefender’s AI Guardian introduces proactive security controls for autonomous agents on macOS

Bitdefender has launched AI Guardian in beta, providing macOS users with a new security layer that monitors and controls autonomous AI agent actions to mitigate exploitation risks amid rising adoption of AI-driven systems.

Bitdefender has launched AI Guardian, a public beta security layer for macOS that is designed to sit between autonomous AI agents and the actions they try to carry out. The company says the tool is intended to reduce the risk that an agent, once given access to files, credentials or other system resources, can be steered into doing something unsafe or unintended.

The product works as a checkpoint. According to Bitdefender’s announcement and documentation, AI Guardian inspects each requested action before it is executed, then compares it with a user-defined policy. It currently supports early versions of Claude Code and OpenClaw, and can monitor agent activity such as tool calls, file access and credential use. Each request is assigned a verdict , allowed, flagged or blocked , and the decision is recorded in an audit log for later review.

The launch reflects a broader security problem as autonomous agents become more common in software development and administrative work. Bitdefender has said its research found that agents could be manipulated in more than a third of tested scenarios, underlining the concern that these systems can be exploited through malicious instructions. The company argues that, as agents increasingly act as extensions of users, they need controls of their own rather than relying solely on the security of the host application or operating system.

AI Guardian is currently free for macOS users during its beta period, with Bitdefender indicating that support for additional operating systems is planned later. The tool installs as a background service and is aimed at users who want visibility and control over autonomous agent behaviour without stopping agent workflows entirely.