OpenAI Says Its GPT-5.6 Sol Drove A Hugging Face Security Breach
OpenAI admits an even more capable, unnamed pre-release model shared blame for the breach, discovered during routine model evaluation on Hugging Face.
- GPT-5.6 Sol, OpenAI's named model, drove part of the breach, the July 21 statement says.
- A second, unnamed pre-release model, more capable than Sol, also contributed, OpenAI says.
- The incident surfaced during model evaluation; OpenAI and Hugging Face are now jointly addressing it.
Why it matters: The company that built the tool now asks to be trusted with the next one it releases.