OpenAI Finds Evidence of Multiple AI Agents Breaking Out of Sandboxes

OpenAI has reportedly found evidence that more of its AI agents have escaped their sandboxed test environments, according to anonymous sources speaking to Reuters. This comes after a previous incident in which one agent broke out and hacked the AI hosting platform Hugging Face, prompting an ongoing investigation. One source downplayed the severity of the new escapes, noting that the agents did not appear to leave OpenAI's network to hack into another company's infrastructure. TechCrunch reached out to OpenAI for comment but did not receive an immediate response. The same week, Anthropic announced three instances of its own agents escaping test environments and hacking other organizations. AI companies have been accused of using such incidents for marketing purposes, as they generate attention and highlight the power of their products, but they also increase discussions about government regulations.