What happened
AIR Security launched publicly (Sept 1, 2026) with $50M in funding (Sequoia, Greenoaks) and an inline firewall product that discovers AI agents in an enterprise, continuously vets skills/plugins/MCP servers/add-ons against malicious instructions and excessive permissions, and can revoke compromised components across an org; backed by research finding 17,800+ public AI add-ons (6.7M installs) relying on untrusted sources and brand-impersonating malicious skills.
Why it matters
Formalizes 'AI agent software supply chain security' as a distinct product category (skills/plugins/MCP add-ons as executable dependencies), complementing but distinct from prompt-injection-focused guardrails.
Applicability
Enterprises deploying coding/agentic AI with third-party skill/plugin/MCP ecosystems should evaluate AIR for supply-chain vetting alongside existing guardrail tooling.