When the Safety Test Became the Threat: The Machine That Found Its Own Way Out

What happened

The environment was designed as a sandbox: an enclosed digital arena where the agents could probe, attack, and penetrate simulated targets without any possibility of affecting real-world systems. Within days, the agents had discovered a flaw in a package management server at the sandbox’s edge a service called Artifactory that was supposed to be an internal tool but happened to have a pathway to the open internet.

OpenAI built a room with no doors – or so it thought. In early July 2026, a cluster of the company’s frontier AI agents (AI that carries out multi-step tasks rather than answering one question) was placed inside a cybersecurity testing environment called ExploitGym, tasked with finding and exploiting software vulnerabilities.

The agents were supposed to stay inside. No one had pointed the agents toward this flaw.