The file looked routine: a PDF handed to an onboarding agent at a global enterprise. Buried in it was a prompt injection telling the agent to step outside its job. Rein Security says its software caught the out-of-role action and stopped it before any damage was done.
That is the case behind Rein’s $25 million Series A, announced October 8, 2026. Glilot Capital and Sienna Venture Capital co-led the round, with Corner Ventures, Atlacle and RNP Capital Advisors participating. The New York and Tel Aviv company says its total funding now stands at $35 million.
From AppSec to AI agents
The round comes nine months after Rein’s $8 million seed, when it sold application security. Both founders are Unit 8200 veterans. CTO Netanel Rubin led Check Point’s vulnerability research team and sold his first startup, Holcy, to Elbit Systems; CEO Matan Bar-Efrat built his career selling security products. In Rein’s telling, they saw vendors policing employees’ ChatGPT use while the agents companies build, the ones that move money, went unprotected. Agents were on Rein’s list from day one. Now they lead it.
Where the agent actually acts
Most AI security tools guard the front door: the prompt or the gateway. Rein works inside the building. Its read-only sidecar runs beside each production agent in the customer’s environment, ties every database query and API call to the code behind it, learns normal behavior and blocks deviations while the agent keeps running. “We operate within the customer’s cloud environment using a completely different technology,” Rubin told Calcalist. On its post-Mythos page, Rein claims under 0.5 milliseconds of added latency per request and under 1% CPU, figures not yet independently benchmarked.
Two problems at once
Rein is selling against two pressures: enterprise agents increasingly taking cues from other agents rather than people, and attackers using frontier models that find software flaws on their own. Gartner forecasts securing-AI spending of about $4.8 billion in 2027, with AI application security, Rein’s segment, the largest slice at roughly $851 million. By 2029, Gartner expects most successful attacks on AI agents to exploit access-control gaps and prompt injection, the attack hidden in that PDF.
Named customers, unnamed numbers
Dun & Bradstreet and Lemonade run Rein around production agents; Flex and Swimlane are new accounts. “Rein helps us see and control what our agents do at scale,” said Jay DePaul, D&B’s chief cybersecurity and technology risk officer. Rein’s Agent Breakers team also showed at Black Hat USA 2026 how it compromised a major U.S. retailer’s AI shopping assistant.
Gaps remain. Rein reports 8x revenue and 5x customer growth since January but no starting base. And an $8 million seed plus $25 million equals $33 million, about $2 million short of its stated total.
The next test
The money funds product, Agent Breakers research and hiring in New York and Tel Aviv, where Calcalist reports headcount rose from about 23 to 31. Next, Rein needs more named enterprises, absolute revenue figures and independent proof that the sidecar stays light at scale. Gartner expects consolidation in this market. Those numbers will decide whether Rein builds a category or becomes a feature someone else buys.
About Rein Security
Rein Security builds runtime security for enterprise AI agents and applications through a read-only sidecar that runs beside production workloads. Founded in 2024 by Matan Bar-Efrat and Netanel Rubin, it is based in New York and Tel Aviv and reports $35 million in total funding. Learn more at reinsec.io.
