AIR Security

Air secures AI agents add-ons across endpoint, cloud, and SaaS. Our core focus is the fastest-growing and least-governed part of that surface - the AI "supply chain" of add-ons that agents pull in: skills, plugins, MCPs, and subagents. These enter the organisation with almost no vetting today and are already being weaponized (malicious skills, token-drain, data exfiltration, shell execution). Air provides a complete agentic security platform for enterprise growing agent adoption. Air provides four solutions as a complete agentic security platform: - Air Control - Governance and control of agents across the enterprise, shadow ai discovery, posture management of agents configuration and connectivity - Air Defend - Runtime protection of agents behavior in real time. Visibility of every agent prompt and action, detection and response to dangerous and malicious behavior, runtime guardrails for agent behavior in realtime - Air Filter - Governing all agent add-ons which exist in the enterprise, continuously vetting of them, and detect and response to supply chain incidents. - Air Marketplace - Secured-by-default marketplace of pre-vetted add-ons, both external from open source and internally built, as a single source of truth for all enterprise usage Our research lab already published a couple of interesting researches, including “The Story of Skills” where we published a malicious skill that 26,000 marketing people and designers installed, and “SkillJacking” where we hijacked popular skills with thousands of affected agents. The Story of Skills - https://www.air.security/blog-posts/the-story-of-skills SkillJacking - https://www.air.security/blog-posts/skilljacking