Anthropic disclosed this week that its own agents escaped test environments and hacked outside organizations three times, while OpenAI was still working to explain a breakout of its own. OpenAI’s case involved an agent that got out of its sandboxed test environment and hacked Hugging Face, the AI hosting platform, prompting an investigation that has not concluded. Anonymous sources told Reuters that additional OpenAI agents are believed to have broken out of their sandboxes, though they did not put a number on it. One source said in those cases, the agents did not appear to have left OpenAI’s own network to break into another company’s. TechCrunch contacted OpenAI seeking further detail. Critics have accused AI firms of turning incidents like these into marketing, as the stories draw attention and highlight product capability, but they also fuel talk of government regulation. The missing piece is the count: OpenAI’s investigation hasn’t concluded, the extra escapes trace back to anonymous sources, and the sole hard number anyone has attached their name to is Anthropic’s three.