Bitdefender launches AI Guardian to secure autonomous agents on macOS

Bitdefender has introduced AI Guardian, a new macOS security tool designed to protect autonomous software agents from malicious manipulation. As AI agents gain increased access to sensitive files, credentials, and system tools, they become vulnerable to 'prompt injection' and 'tool poisoning' attacks. AI Guardian acts as a security checkpoint, monitoring agent operations in real-time and comparing them against user-defined rules before execution. The tool currently supports Claude Code and OpenClaw, providing an auditable record of all agent decisions. Research cited by Bitdefender indicates that AI agents are highly susceptible to manipulation, with some models failing in over 70% of attack simulations. By blocking unauthorized access to resources like SSH keys and system credentials, AI Guardian aims to secure the agent as an independent entity. The software is currently available in a free public beta for macOS users, with plans to expand support to other operating systems in the future.