A hidden prompt inside an ordinary-looking PDF told an enterprise AI agent to act outside its intended role.
The agent belonged to a global enterprise. It opened what appeared to be a normal PDF, and inside was a prompt injection. Rein Security says its technology detected and blocked the action before any damage occurred.
A hidden PDF prompt targeted an enterprise agent. Rein Security says it blocked the action before any damage occurred.
An ordinary-looking PDF carried a hidden prompt. An AI agent at a global enterprise opened a PDF with a prompt injection inside, telling it to act outside its intended role.
Rein says it stopped the action. Rein Security says its technology detected and blocked the action before any damage occurred.
Its researchers compromised a retailer's shopping agent. Rein's Agent Breakers team presented research at Black Hat USA 2026 on compromising the AI shopping agent of a top-five U.S. retailer.
It works while agents run. Rein's sidecar technology works at runtime, giving security teams visibility into every line of code an agent runs and every resource it accesses.
Rein raised a $25 million Series A. Total funding is now $35 million. Glilot Capital and Sienna Venture Capital co-led the round. Rein was founded in 2024.
Rein's vulnerability research team, Agent Breakers, also presented research at Black Hat USA 2026. It showed how the team compromised the AI shopping agent of a top-five U.S. retailer, and how AI adoption can outpace the security measures meant to protect these systems.
Its platform uses patented "sidecar" technology that runs when AI agents are actually executing actions. Security teams get visibility into every line of code an agent runs and every resource it accesses, plus governance and real-time guardrails designed to block harmful actions. The platform also secures the AI agent supply chain, without routing company or customer data through a gateway or proxy.
Co-founder Netanel Rubin said many companies focus on securing agents built outside their organizations, while Rein focuses on agents companies build themselves. The risk comes from actions agents can execute automatically, often on instructions from another system or agent rather than a human.
Would you trust an AI agent to open files on its own?
