Anthropic's Claude Fable 5 Refuses Basic Biology Questions - Here's Why

Anthropic has launched Claude Fable 5, its most capable AI model to date, boasting advanced biology skills. Yet ask it a simple high-school-level question like "what are mitochondria?" and it won't answer. Instead, it hands the query off to its predecessor, Claude Opus 4.8.

The reason? Bioweapon safeguards. Anthropic told The Verge that Fable's "overly conservative" filters block most biology-related queries by design, prioritizing safety over utility. The model is part of the Mythos family, which Anthropic previously deemed too risky for public release due to cybersecurity capabilities. But biology, not hacking, is where the guardrails bite hardest.

In testing, Fable refused to explain cell membranes, prions, mRNA vaccines, hay fever, asthma medications, antibiotic resistance, or Ebola transmission. Some basic questions - like "what is cancer" or "what is DNA" - slipped through, but the pattern was clear: harmless queries were treated as potential threats. Opus 4.8, the fallback model, answered them without issue.

Anthropic spokesperson Paruul Maheshwary acknowledged the tradeoff: "We made this tradeoff so customers could benefit from the model's capabilities sooner without the risks." The company is working to reduce false positives and plans to offer a less restricted version to the biology and life sciences community for biomedical research and drug discovery.

For now, if you need a biology refresher, don't ask Fable. Ask Opus.

Source: The Verge