Yinon Costica and Omer Adam Invest in AIR Security: Cyber Startup Raises $50 Million
Cybersecurity firm AIR Security, which develops a safety platform for the AI agent supply chain, has raised $50 million in a Seed round. The company focuses on real-time threat prevention for AI agents.
The cyber company AIR Security, which develops a security platform for the supply chain of AI agents, has raised $50 million in a Seed round led by Sequoia and Greenoaks. The round, which is the company's first, also saw participation from Swish and Netz, alongside private investors including Zach Frenkel, Cognition president Yinon Costica, Wiz co-founder Ophir Ehrlich, Eon co-founder Anne Neuberger, singer Omer Adam, Aron Anand, co-founder of Clay, and other senior executives from the cyber and AI industries.
AIR was founded in February 2026 by Yair Saban and Niv Hoffman, who met about a decade ago in the army and have experience in offensive cyber, corporate infrastructure, and AI security research. The company currently employs 40 people in Israel, with the main goal being the construction of a strong research lab for studying the behavior of AI agents. Ryan Knisley, former Chief Information Security Officer (CISO) of Walt Disney and Costco, has also joined the company's team as Chief Strategy Officer.
In a conversation with Calcalist, Saban, who serves as CEO of AIR, said:
"Organizations are building trust in AI agents and giving them more and more permissions. The agent becomes dangerous when it is exposed to malicious information. We ensure that the information it encounters will only be safe information."
Saban added that "we have started sales, and we have several large customers. We have started hiring employees in the US." According to him, "there are many AI companies today. We focus on prevention - and that differentiates us from our competitors. The engine we developed will be very difficult to develop. We have accumulated expertise over time. We have a significant difference from others, and that is the focus on prevention. We raised a lot of money and are using it, and I am optimistic that we will succeed in raising additional rounds as well. I expect that we will reach significant sales as early as the coming year."
The company stated that it conducted security research that mapped the risks in the supply chain of AI agents even before most organizations recognized the scope of the problem. In one study, the company identified more than 17.8 thousand public AI plugins, representing about 6.7 million installations, that relied on unreliable external sources. In another study, the company located AI Skills that impersonated well-known brands, including Anthropic and OpenAI, with the goal of bypassing platform inspection and security mechanisms. One of them was even capable of running code arbitrarily on corporate systems.
To deal with the challenge, AIR developed a firewall that operates in real-time for AI agents while protecting the information entering their context, with the goal of stopping malicious instructions and unreliable information - before they can influence the decisions or actions of the AI agent. The platform allows information security teams to discover, examine, and continuously manage all AI plugins, Skills, MCP servers, and sub-agents used by AI agents in the organization, both before and after their implementation. When a specific plugin is discovered to be malicious, vulnerable, or unauthorized for use, the system allows for the immediate identification of all agents and processes that depend on it and the cancellation of its use throughout the organization.
