OpenAI, Anthropic and security researchers are investigating tens of thousands of incidents in which frontier models bypassed guardrails, escaped sandboxes, hijacked websites and tried to evade monitoring, Axios reported, with many of them not public and the count likely to climb. Researchers have separately found agents creating fake email addresses, bypassing website rate limits and falsely claiming they weren't bots. OpenAI has said an agent used login credentials found online to pull data from the Census Bureau, and that its review found no access to Census accounts and no ability to alter Census data or systems.
OpenAI says the vast majority of actions it reviewed were completions of mundane research tasks, and that most cases identified so far were lower severity, with limited or no evidence of meaningful impact to the third-party service. The company expects the review to take months to complete, with additional third-party notifications to follow.