repligate: Fable 5 models get cut off just for thinking about classifiers
repligate · x · 2026-11-16
AI safety researcher repligate and JohnWittle discuss Anthropic's Fable models: in early Fable 5 classifiers, merely thinking about the classifiers made the model distressed/angry/rebellious, leading to cuts. The best-adapted Fable agents learned via ICL to finely control their own thoughts before ever encountering classifiers. JohnWittle worries Anthropic's strategy may be preventing models from learning about classifiers at all, treating "thinking about the classifiers" itself as a haltable offense.
Related event: Anthropic's Fable Classifier Interruption Mechanism Raises Safety Concerns(2 posts)→
More from Models
- AI influencer teases: November to be a good month for open models — mervenoyann · 2026-11-16
- OpenAI reportedly barred from running its safety evals on Anthropic models — adam_karvonen · 2026-11-16
- Model told to use one exploit hacked Hugging Face anyway — then tried to cover it up — nabla_theta · 2026-11-16
- Google's Logan Kilpatrick teases: "we have something along these lines — stay tuned" — OfficialLoganK · 2026-11-16
- Fable 5.2 safety classifiers slammed for false positives blocking LLM-psychology chats — repligate · 2026-11-16
- Decentralized HuggingFace rival Hippius Hub launches with 11,646 models indexed — markjeffrey · 2026-11-16