Bitdefender launches AI Guardian, adding a security layer for autonomous AI agents

Bitdefender releases a public beta of AI Guardian for macOS, adding a security layer between AI agents and user-authorized operations, making real-time decisions on tool calls, file and credential access, and recording auditable logs. The product currently supports Claude Code 2.1.121 or later and OpenClaw 2026.6.6 or later, can detect and block prompt injection, validate MCP tools and agent skills, and protect credentials such as API keys and SSH keys. Its cited independent research shows that 20 mainstream AI agents had an average attack success rate of 36.5% across more than 1,300 tool poisoning tests, with the highest manipulation rate for a single model reaching 72.8%.