@haydenfield Yeah exactly. It's not that the model is too powerful but the specific issues in question aren't Fable-specific and the specific tasks can also be solved by weaker models in this case. Ie there is no narrative violation by Anthropic here. Maybe I'm too pedantic ofc 😅
Hayden Field@haydenfield
@BlackHC It's very apparent via both internal sources and public statements; the company is trying to make the point over and over that the jailbreak in question and the issues in the white paper could be achieved by OpenAI's GPT-5.5 and other models. Fable 5 specifically.
6:25 AM · Jun 16, 2026 · 43 Views