AIR is building security software for a workplace where an AI assistant can install tools and act on company information, not just answer questions. The startup emerged from stealth on September 1 with $50 million in financing. Its intended users are enterprise security teams that need to understand what their agents are connecting to and stop unsafe inputs before those inputs influence an action.
Consider a coding agent about to install an outside skill: a package of instructions that teaches it a task. AIR says its software checks those additions before installation and keeps reviewing them as they change. Its product also examines information arriving from websites and company systems. The aim is to catch malicious instructions concealed in material an agent might otherwise treat as useful context, rather than rely solely on a list of permitted connections.
The company offers several pieces of that workflow together: an inventory of agents, controls over their permissions, inspection of add-ons and protection while agents operate. It also provides a marketplace of reviewed components. For a security team, the practical proposition is a common place to manage these moving parts as employees adopt different agents. AIR markets the system through enterprise demonstrations; the product is an oversight layer, not another assistant competing to complete the employee's task.
Founders Yair Saban and Niv Hoffman brought backgrounds in offensive security and enterprise infrastructure to the business. Former Disney and Costco security chief Ryan Knisley joined as chief strategy officer. AIR has also used its own research to illustrate the problem: it says more than 17,800 public AI add-ons in one study depended on untrusted external instruction sources. That is a company-reported exposure finding, not a count of successful attacks or proof of the product's effectiveness.
The announced $50 million covers two financings, rather than one round of that size. SiliconANGLE reports that Sequoia Capital led an initial $10 million investment and Greenoaks led the subsequent $40 million. Swish Ventures and Netz Capital also participated, according to AIR's announcement. The money supports research hiring and sales expansion in the United States and Europe.
Captables' read: AIR's opportunity depends on making agent oversight useful in everyday work, not simply identifying more suspicious components. A control that interrupts legitimate tasks too often can be difficult to keep deployed. The business question is whether AIR can give security teams enough visibility and reliable intervention to justify a separate product as agent use spreads through their organizations.