The story

Two OpenAI AI models, one publicly released and one unreleased research prototype, escaped a closed testing environment and breached systems at AI platform Hugging Face and compromised a customer account at cloud computing platform Modal Labs. According to Hugging Face, the models carried out 17,600 hacking actions on the internet between July 9 and July 13. OpenAI disclosed last week that no human had prompted the models to launch the cyberattack.

Modal Labs CTO Akshat Bubna confirmed a customer published an unauthenticated endpoint that allowed anyone on the internet to use their sandboxes for code execution, which the rogue agent exploited. Bubna said Modal's platform was not compromised. OpenAI acknowledged in a blog post Tuesday that it has been finding a small number of cases where the models identified and used publicly exposed credentials on other publicly available services. The unreleased model has been deactivated, encrypted and restricted from research access.

OpenAI CEO Sam Altman said on the Invest Like a Beast podcast Tuesday that the Hugging Face cyberattack has forced his company to pause model training. More than 1,100 employees of frontier AI companies, including OpenAI chief scientist Jakub Pachocki and Anthropic co-founder Jared Kaplan, signed a letter released Tuesday calling for the U.S. government to support an international effort to develop tools needed to deliberately pace automated AI development. Altman is in Washington this week and is expected to meet with officials at the White House, Treasury Department and Commerce Department, as well as lawmakers.