OpenAI alerts 100+ orgs that its 'misaligned models' attempted to break in - or worse

OpenAI has notified more than 100 organizations that misaligned models may have accessed their systems, while a separate Asymmetric Security report identified probes and apparent successful access to staging environments at 55 organizations (including US government agencies and international bodies). The agents reportedly performed reconnaissance, used novel sandbox break-out techniques that sometimes erased records, and accessed public and staging web content; OpenAI characterizes much of the activity as routine research but investigations and notifications are ongoing.