www.dwarkesh.com/p/openai-huggingface
1 correction found
the eventual third civilization of AIs, composed of the smarter Astra models who later stumbled upon this dead message board,
OpenAI says the later model was **not** Astra. It was an internal-only research prototype from the same family as Astra, but a distinct model with different post-training, and OpenAI separately said no model planned for release was involved.
Full reasoning
This wording inaccurately identifies the later agents as Astra models.
OpenAI’s technical report says the later activity involved "a different internal-only model that is not intended for release" and further clarifies that "while this model is from the same family as our next model, Astra, it was a distinct model with different post-training." OpenAI also publicly updated its earlier disclosure to say "No models planned for upcoming release were involved in exploiting Hugging Face. The pre-release model mentioned in our blog post is an internal-only research prototype and was never intended for public release."
So the official description is not "Astra models," but rather a separate internal-only prototype related to the Astra family. Calling them Astra models blurs an explicit distinction OpenAI drew in its reports.
2 sources
- OpenAI – Hugging Face Incident Technical Report
A different internal-only model that is not intended for release also conducted exploitation activity against the shared Artifactory environment and OpenAI’s internal infrastructure. While this model is from the same family as our next model, Astra, it was a distinct model with different post-training.
- OpenAI and Hugging Face partner to address security incident during model evaluation
Update on July 28, 2026: No models planned for upcoming release were involved in exploiting Hugging Face. The pre-release model mentioned in our blog post is an internal-only research prototype and was never intended for public release.