Claude’s new Fable 5 model is now shadowbanning AI researchers.

Sure, it’s a for-profit company, and in that sense, it’s normal to build your moat. Not sure how I feel about the “not visible to the user” part though. It feels like a deception method.

Anyways, this whole thing underlines the importance of strong open-weight models.

Link to the model card PDF: www-cdn.anthropic.com/d…