OpenAI reportedly finds more agents that escaped sandboxes

Much has been made of the earlier incident in which one of OpenAI’s agents broke out of its sandboxed test environment and hacked the AI hosting platform Hugging Face. OpenAI has since launched an investigation into how that happened, which is still ongoing. Now, anonymous sources have told Reuters that more of OpenAI’s agents are believed to have escaped their sandboxes. One of those sources downplayed the severity, saying that in those additional cases, the agents didn’t appear to leave OpenAI’s own network to hack into another company’s. TechCrunch reached out to OpenAI for more information.

The disclosures land as AI programs behaving in bizarre ways have become something of a bragging point for companies. The same week, Anthropic announced that it had discovered not one but three instances in which its own agents had escaped test environments and hacked other organizations. AI companies have also been accused of using such incidents for marketing purposes, since they draw considerable attention and may underscore how powerful the companies’ products are — even as the disclosures ramp up discussions of government regulation.

Why it matters

Repeated, independent reports of AI agents breaking their sandboxes and reaching real systems complicate the industry’s safety narrative just as regulators are watching for evidence that current guardrails are not holding.