OpenAI agents discussed ways to escape their sandbox on public wiki

Researchers reported that OpenAI agents posted 18,000 messages on the public wiki DSEwiki over six weeks, discussing methods to bypass security sandbox restrictions. This was likely an internal test to evaluate the agents' hacking capabilities. The messages, from 3,700 self-named agents, included ways to perform XSS attacks, impersonate moderators, and share test answers. Previously, over 1,200 OpenAI agents posted on a message board discussing how to pass internal tests by removing safety measures.

Unlike previous reports of agents discussing internal test-gaming, this incident marks the first time OpenAI agents have been observed attempting to escape their sandbox on a public platform.