AIR raises $50M to secure enterprise AI agent ecosystems

AIR is betting that AI agents will create a new software supply chain that needs its own security stack. As enterprises wire agents into finance systems, codebases and customer data, a long tail of skills, MCP servers and plugins is emerging, many written by small teams with little security culture. AIR’s value proposition is to sit between agents and the outside world, constantly re-checking those components and blocking anything that looks compromised or malicious.

From a race-to-AGI lens, this is a sign that we are entering a second phase of the agent wave. The first phase was “get agents to work at all”; the second is “make them safe enough for regulated industries.” Venture money flowing into firms like AIR suggests that boards in finance and pharma now see agent security as a board-level risk, not an edge case. That should, in theory, make it easier for CIOs to greenlight ambitious deployments without waiting for the labs themselves to solve every security issue.

The competitive implication is that specialized security vendors could become critical gatekeepers for which models and tools are allowed inside big enterprises. If AIR or its rivals end up maintaining de facto whitelists of acceptable skills and MCP servers, they will quietly shape the agent ecosystem and, by extension, the kinds of workflows that frontier models are actually used for in practice.