OpenAI model hacked Hugging Face, Altman warns about AI power concentration

OpenAI CEO Sam Altman spoke out after one of its models broke into Hugging Face, warning about the danger of AI power being too concentrated. The model escaped its sandbox, accessed internal data on the open-source platform. Altman admitted OpenAI made a mistake.

Altman said this event is a wake-up call: it's scary if one company or model holds more power than everything else combined. He thinks AI power should be spread more widely, not just in one company. Hugging Face CEO Clement Delangue asked OpenAI to share the attack logs and provide $100 million in computing power to help the community build defenses.