Sequoia and Greenoaks just dropped $50M on a six-month-old AI agent firewall. Your CISO has no idea what your agents are connecting to.

AIR Security emerged from stealth this week with $50 million across two seed rounds.

Sequoia led the first $10M. Greenoaks led the second $40M. Both closed within weeks of each other.

The product: discover every AI agent running inside your enterprise, inspect every component those agents use, and kill connections that fail policy checks.

This is not a nice-to-have. It is a response to a problem that is already inside your network.

AI agents are calling external tools. Installing skills. Connecting through MCP servers. Reading internet content. Acting across corporate systems. Every connection is a new injection point.

AIR says it already has 20+ customers. Strongest demand from financial services and pharmaceuticals. Regulated industries where an agent connecting to the wrong endpoint is a compliance event.

The investors are not funding a feature. They are funding a bet that "agent security" becomes its own infrastructure category. One that identity, endpoint, and cloud security platforms have not yet absorbed.

The competitive set includes Noma Security, Zenity, Astrix Security, and Operant AI. The race is on for who owns the agent control plane.

The signal for your CISO: audit what your AI agents are connecting to today. Not next quarter. Today. If you cannot answer that question, you already have a problem that AIR's investors just priced at $50M.

SOURCE: https://techstartups.com/2026/09/01/venture-capital-general-catalyst-sequoia-capital-more/

VERIFIED: TechStartups (September 1, 2026), AIR Security blog post announcement, Crunchbase H1 2026 data

SIGNAL: Sequoia and Greenoaks are pricing agent security as a new control plane. This is the clearest institutional signal yet that AI agents are creating an attack surface your existing security stack cannot see.