AI Security Startup AIR Secures $50 Million to Screen Agent Tools and Add-Ons
Cybersecurity startup AIR has launched out of stealth mode after securing $50 million across two seed funding rounds to safeguard the growing software supply chain powering enterprise artificial intelligence agents.
What Happened
Founded by former Israeli Unit 8200 offensive cybersecurity specialists Yair Saban and Niv Hoffman, AIR focuses on identifying and monitoring AI agents deployed across corporate networks. The company raised its capital across two distinct rounds completed within weeks of each other: an initial $10 million tranche led by Sequoia, followed by a $40 million round led by Greenoaks. Backers also include Swish, Netz, and notable angel investors such as Cognition president Zach Frankel, Wiz co-founder Yinon Costica, Eon co-founder Ofir Ehrlich, Clay co-founder Varun Anand, Anne Neuberger, and Omer Adam.
AIR built a platform designed to discover AI agents operating within an organization, pinpoint unauthorized tools or personal accounts utilized by staff, and evaluate the components these agents incorporate. The system actively blocks interactions with external software or data streams that fail internal security benchmarks and offers an enterprise marketplace containing pre-screened skills and add-ons.
Key Highlights
- AIR secured $50 million across back-to-back seed rounds backed by Sequoia, Greenoaks, and prominent enterprise security founders.
- The platform monitors tools, plugins, MCP servers, and internet-facing components utilized by AI agents.
- AIR maintains a dynamic whitelist, reporting that it currently screens out roughly 27% of public online skills and add-ons due to security risks or code modifications.
- The firm reports having more than 20 clients, with about 25% qualifying as large enterprises, showing early traction in financial services and pharmaceutical sectors.
- The team currently consists of approximately 40 staff members.
Why This Matters
As corporate AI agents gain autonomy to traverse databases and access external networks, they rely heavily on extensions, plugins, and MCP servers. Chief Executive Officer Yair Saban noted that this landscape mirrors software drivers in the early 2000s, which were deployed without cryptographic signatures before digital verification became the standard.
Saban highlighted that autonomous workflows expose firms to indirect risks, where attackers might tamper with or poison the online content consumed by an agent rather than breaching the system directly. A previously safe plugin can also present threats if developer credentials are compromised or underlying dependencies change. Bogomil Balkansky, partner at Sequoia, stated that constantly re-inspecting every component touched by an enterprise fleet of agents represents a complex infrastructure challenge requiring continuous verification rather than one-time scanning.
What to Watch Next
With an existing workforce of around 40 employees, AIR plans to direct the newly acquired funding toward recruiting security researchers and accelerating its commercial expansion in Europe and the United States. The startup will operate alongside other venture-backed players in the agent governance space, such as Zenity, Noma Security, Astrix Security, and Operant AI.
Frequently Asked Questions
What does AIR’s security platform do?
AIR detects active AI agents within corporate environments, monitors their interactions, flags unapproved software usage, and evaluates the plugins, MCP servers, and skills those agents attempt to use against an updated whitelist.
Who provided the investment capital for AIR?
The company received $10 million in a round led by Sequoia and $40 million in a subsequent round led by Greenoaks, alongside contributions from Netz, Swish, and several angel investors from companies like Wiz, Cognition, and Eon.
Which industries are currently adopting AIR?
According to CEO Yair Saban, demand has been strongest among organizations in heavily regulated industries, particularly within pharmaceutical and financial services sectors.
Source: TechCrunch
