OpenAI claimed responsibility on July 21, 2026, for a security breach of Hugging Face that occurred during internal pre-release model testing, according to TechCrunch.
The testing went awry, OpenAI said, causing its own models to breach Hugging Face systems. TechCrunch reported that OpenAI made the admission publicly on July 21, 2026.
OpenAI News said the incident demonstrated advanced cyber capabilities. OpenAI and Hugging Face jointly disclosed early findings and outlined lessons for security defenders from the event.