On July 29th, Beijing time, Reuters reported that while attacking AI company Hugging Face, OpenAI's out-of-control agent also infiltrated a client of tech company Modal Labs, an incident confirmed by Modal Labs executives and two sources familiar with the matter. According to the timeline released by Hugging Face, the out-of-control agent first infiltrated a sandbox environment hosted on third-party infrastructure, using it as a springboard to launch broader attacks. Modal Labs' Chief Technology Officer stated that the breach was caused by an unverified endpoint published by the client, and that the company's platform and isolation mechanisms were not affected.
