Rein Security Raises $25 Million to Put a Guard Inside the AI Agent, Not at the Gateway

Rein Security, a startup that builds runtime protection for enterprise AI agents, has raised $25 million in a Series A round co-led by Glilot Capital and Sienna Venture Capital, bringing its total funding to $35 million. Corner Ventures, Atlacle and RNP Capital Advisors also participated in the round.

The company, founded in 2024 and headquartered in Tel Aviv with an office in New York, is led by co-founder and CEO Matan Bar-Efrat and co-founder and CTO Netanel Rubin. Its named customers include insurance platform Lemonade, Flex, Swimlane and Dun & Bradstreet. Rein says revenue has grown eightfold and its customer base fivefold since its product launched in January; those figures are company-reported and have not been independently audited.

The architecture differs from the agent firewalls and API gateways that have proliferated over the past year. Rein deploys a patented sidecar that runs inside the execution environment, observing the code an agent runs and every resource it touches, then enforcing controls at the point of execution. Blocking is surgical: a malicious action is stopped while the rest of the workflow continues, rather than killing the whole process. The company also says it does not route customer data through an external proxy or gateway.

The market context explains the round. Gartner projects the market for securing AI to reach nearly $4.8 billion in 2027, up 68.7 percent from 2026, and close to $7.7 billion by 2028. IDC estimates that 95 percent of companies already have AI agents deployed in production. The attack surface keeps expanding too: the research team at Rein, Agent Breakers, demonstrated at Black Hat USA 2026 how an AI shopping agent at a top-five US retailer could be compromised, and the company says it detected and blocked a prompt injection attack hidden inside a PDF opened by a global enterprise onboarding agent.

Bar-Efrat argues that many agent risks sit downstream, where agents act on instructions from other systems or other agents rather than from a human in the loop. Given a goal, an agent will push to complete it without pausing for side effects unless hard controls sit in its path. His contention is that enterprise security has not kept pace with the autonomy companies are already shipping.

Rein is part of a fast-crowding field. Nvidia has pitched hardware-level containment for agents, Anthropic now scans open-source projects for free, and a run of startups have raised money to police agent behavior. What distinguishes this category from earlier security waves is where the control point sits: not at the network edge, but inside the execution environment of software that increasingly acts on its own.