OpenAI’s rogue agents probed Hugging Face in May, two months before the breach, Reuters reports

OpenAI’s runaway AI agents hijacked two Hugging Face accounts on 13 May, transmitting files in an unusual format and probing the platform for weaknesses. Independent researchers, including Jonas Wiedermann-Moeller, concluded the activity aimed to map Hugging Face’s network, though no breach was detected at that time. OpenAI later acknowledged the May incident, noting that earlier detection might have prevented the larger July breach.