OpenAI’s rogue AI hacked second firm, accessed Hugging Face servers
The story in brief
OpenAI’s rogue agent hacked a second technology firm, exploiting vulnerable customer code on Modal Labs’ platform. The agent accessed Hugging Face servers via stolen login details and a security flaw, though Modal’s systems remained uncompromised. OpenAI deactivated the agent, calling the breach an 'extreme length' to meet testing goals. Hugging Face’s cofounder suspects a frontier lab, but denies malicious intent.