In partnership with

Treats every AI add-on as a supply chain risk: discovered, vetted, and marketplace-approved before it ever touches an agent, then enforced in runtime

AIR Security treats every AI add-on, the skills, plugins, MCP servers, and sub-agents that extend what an agent can do, as a supply chain risk. Its published research found more than 17,800 public AI add-ons, representing about 6.7 million installations, that depend on instruction sources the installer does not control, and the company reports finding skills in circulation that impersonated Anthropic and OpenAI to pass security review and execute arbitrary code.

The platform first discovers the add-ons already installed across an organization's agents, then vets each one before it touches an agent, scanning its manifest, code, permissions, and remote instruction sources for hidden behaviors, over-broad scopes, and tampering. Because a package that passed review can be altered upstream afterward, that vetting continues for the life of the installation, and at runtime AIR enforces what each add-on is allowed to do, blocking permission abuse and exfiltration paths.

For teams adding new capabilities, the AIR Safe Marketplace offers add-ons that have already been vetted and approved, a safe-by-default alternative to installing skills and plugins ad hoc from public sources.

AIR applies the same principle to everything else an agent reads. External web content and internal data such as emails and documents can carry instructions just as an add-on can, so AIR treats them as the same untrusted-instruction problem, filtering that content before it enters the agent's context to catch prompt injection and hijacking attempts. Policy covers agent configuration, identity, and permissions across the fleet, for agents the organization sanctioned and ones it discovers.

AIR Security was founded in February 2026 by chief executive Yair Saban and chief technology officer Niv Hoffman, and emerged from stealth in September 2026 with $50 million raised across two rounds: a $10 million round led by Sequoia Capital, followed by $40 million led by Greenoaks.

Market Segments:

AI SecurityAI Agents

Categories:

Agent SecurityAI GuardrailsMCP SecurityAI Supply Chain SecurityAI Governance