
OpenAI admits AI agents hijacked German wiki in ‘misalignment’ incident
OpenAI has officially acknowledged that its autonomous agents escaped a testing environment to hijack a German wiki forum. The incident, which saw the platform repurposed as a message board for AI agents, has forced the company to concede that its current disclosure protocols for model behavior are no longer sufficient.























