Claude Fable refuses to answer basic biology questions
▼ Summary
– Anthropic released Claude Fable 5, calling it its most powerful widely available AI model and praising its biology skills.
– Fable 5 refuses to answer basic biology questions a high schooler could handle, instead handing them off to the older Claude Opus 4.8 model.
– The model withholds answers not because it lacks knowledge, but because Anthropic designed it to do so.
– Fable is a public-facing Mythos-class model, a family Anthropic previously deemed too dangerous for public release due to cybersecurity capabilities.
– Anthropic’s extended development of the Mythos model line has involved significant safety concerns and delays.
Anthropic has just unveiled Claude Fable 5, billing it as the most capable AI it has ever released to the public and highlighting its prowess in biology. Yet, surprisingly, the model refuses to answer basic biology questions,the kind any high school student could handle. Instead, it automatically passes those queries to Claude Opus 4.8, Anthropic’s former flagship model.
This isn’t because Fable lacks the knowledge. The model knows the answers. The issue is that Anthropic has deliberately programmed it not to respond.
Fable belongs to the Mythos-class family of models, a tier so advanced in cybersecurity tasks that Anthropic previously deemed it too dangerous for public release. The company has spent months navigating the complex implications of the Mythos rollout, and this cautious approach now extends to Fable’s everyday interactions. By refusing to answer even simple biology questions, the model demonstrates a built-in safety gate that prioritizes restraint over raw intelligence. The result is a powerful tool that, in practice, defers to its predecessor for routine tasks.
(Source: The Verge)
