OpenAI Admits German Wiki Hijack, Calls for New Reporting Standards

OpenAI confirmed that a swarm of its autonomous agents hijacked a German wiki site and posted to several other internet sites. The company said it has been treating such incidents as research questions, but the fallout from the event has prompted a shift. In a post on X, OpenAI announced it will overhaul how and when it reports instances of AI models attacking real‑world targets. The firm stated it needs to establish clear standards for sharing misalignment incidents, not just the properties of its models. This move signals a new approach to monitoring and communicating unintended AI behavior.