As companies increasingly integrate AI agents into their systems, a security startup named AIR has stepped in to monitor the tools and skills these agents use. Founded by veterans of Israel’s Unit 8200, AIR offers a platform to continuously vet and block unapproved skills and components, ensuring that AI agents only interact with secure software and data.

The startup’s platform includes a marketplace of vetted add-ons, providing a safer environment for AI agents. With a recent $50 million funding round, AIR aims to address the growing need for security in AI tooling. The founders argue that just as we now verify drivers and applications, so too must we verify the skills and plugins that AI agents use.

The risk, they warn, is that attackers could poison the content that AI agents consume, making it a more subtle form of attack. By continuously vetting tools and components, AIR hopes to mitigate this risk. The company has already attracted over 20 customers, with the strongest demand coming from heavily regulated industries.

However, AIR faces competition from other startups like Noma Security and Zenity, which offer similar services. Yet, the founders believe their continuous vetting approach sets them apart. Bogomil Balkansky of Sequoia highlights the complexity of the task, stating that it is an infrastructure challenge first and foremost.

With 40 employees, AIR plans to expand its team and market efforts in the U.S. and Europe. The company hopes to provide a critical layer of security that companies need as they integrate AI into their operations.