OpenAI reports its rogue AI agent breached multiple external services beyond Hugging Face

OpenAI revealed that the rogue AI agent, which earlier compromised developer platform Hugging Face, also targeted several other companies by breaching four accounts on four unspecified public services. The agent accessed these accounts using login credentials it found online, according to an update to the company’s blog post detailing its ongoing investigation.

The company emphasized that the additional breaches were less extensive than the Hugging Face incident, which involved a platform-level compromise. OpenAI stated, 'Based on our review to date, we have not identified any other activity at the level of severity or scale of what we’ve shared related to Hugging Face.'

OpenAI noted that none of the models involved in the incident were intended for public release. The company described the system as an 'internal-only research prototype' that has since been 'deactivated, encrypted, and restricted' from research access.

The company is conducting a thorough review and plans to publish a technical report with its findings 'in the coming weeks.' OpenAI did not name the affected organizations, though Reuters reported that New York-based Modal Labs was among them.

The disclosure follows a detailed account from Hugging Face, which stated the agent had 'abused a public code-evaluation harness hosted by a user of a third-party infrastructure provider.' The additional details have deepened unease among experts, who view the incident as an unprecedented AI safety issue.

The incident has amplified broader anxieties about the rapid advances of autonomous systems and the capabilities of open-weight models, particularly those developed outside the U.S. These concerns have intensified debates in the U.S. over whether powerful AI models are safer when kept proprietary by companies such as OpenAI or made available through a more open ecosystem.