AIR Security launches with $50M to build a firewall for AI agents

Artificial intelligence agent security startup AIR Security Inc. formally launched today with $50 million in funding to build what it calls an inline firewall for AI agents.

Founded earlier this year, AIR is going after a gap that has opened as coding agents spread through large enterprises. Those agents reach for third-party skills, plugins and Model Context Protocol servers on their own initiative. The security team rarely sees what got picked up. AIR’s software sits in that path and screens the instructions, tools and data heading into an agent’s context before the agent acts on any of it.

Discovery comes first. The platform maps the agents already running across endpoints, cloud accounts and software-as-a-service applications, then keeps re-checking the components those agents depend on. A skill that passed review in March can be rewritten in June by whoever maintains it. Untrusted tools and injected instructions are blocked at runtime, and AIR also runs a marketplace of add-ons it has already cleared.

Company researchers have put some numbers on the exposure. More than 17,800 public AI add-ons, accounting for 6.7 million installations, drew instructions from outside sources nobody had verified. The team also turned up AI skills impersonating Anthropic PBC and OpenAI Group PBC that could run arbitrary code once installed. Roughly 27% of the add-ons and skills AIR finds online get filtered out.