AIR Security has formally launched with $50 million in funding, raised across two rounds in its first six months, to build an inline firewall that protects AI agents from untrusted tools and instructions. The company says it found more than 17,800 public AI add-ons pulling instructions from unverified sources, with roughly 27% flagged as potentially dangerous, including skills impersonating trusted brands like Anthropic and OpenAI. AIR’s platform continuously monitors and re-verifies the skills, plugins, and MCP servers an organization’s AI agents depend on, and already serves more than 20 companies in sectors such as financial services and pharmaceuticals.
