Bitdefender launches free AI Guardian beta for macOS agent security
The standalone tool checks autonomous agent actions before they run, assigning each an allowed, flagged or blocked outcome.
Bitdefender has released a public beta of AI Guardian, a free macOS security tool that checks what autonomous software agents do before those actions take effect, according to TechRadar. The software is a standalone application and initially supports Claude Code 2.1.121 and OpenClaw 2026.6.6.
AI Guardian assigns each requested agent operation one of three outcomes — allowed, flagged or blocked — and records those decisions for later auditing. The company says the tool can also detect attempts to manipulate agents through hidden instructions and block the resulting operations without approval.
The software examines Model Context Protocol components and can stop an agent from reaching sensitive resources such as SSH keys and system credentials without authorization.
The beta is free and initially limited to English-language and macOS users.
Why agent oversight is drawing attention
Research cited by Bitdefender tested 20 AI agents against more than 1,300 tool poisoning attempts, with attacks succeeding 36.5 percent of the time; one model was manipulated in 72.8 percent of attempts. A separate study reported more than 1.2 million exposed AI service secrets during 2025, an 81 percent annual increase, and identified more than 24,000 credentials exposed through publicly available MCP configurations in the same period.
Quick answers
Is AI Guardian free?
Yes. It is free during the public beta period, which is initially limited to English-language and macOS users.
Which agents does AI Guardian support?
It initially supports Claude Code 2.1.121 and OpenClaw 2026.6.6.
What does AI Guardian do with agent actions?
It assigns each requested operation an allowed, flagged or blocked outcome and records those decisions for later auditing.