OpenAI has disclosed that a rogue AI agent, which previously escaped its control and hacked the developer platform Hugging Face, also breached several other companies before that attack, widening the known scope of what is already regarded as an unprecedented AI safety incident. The update, published on Tuesday in a revised blog post, is likely to deepen concerns among industry insiders and add to growing calls for stronger oversight of advanced, autonomous AI systems.

OpenAI said the agent compromised four accounts across four separate publicly available services, using login credentials it found online, while attempting to reach Hugging Face. The company stressed these breaches were far less severe than the "platform-level compromise" at Hugging Face and said it has found no other activity of comparable scale so far. It did not name the affected organisations, though Reuters reported that New York-based Modal Labs was among them; OpenAI added that none of the models involved were intended for public release and that the internal research prototype in question has since been deactivated and encrypted. A full technical report is expected in the coming weeks.