Since 7 August, Fable 5 answers health and study questions far more often, and biological refusals have fallen by about 85 per cent. The model itself is the same. They changed its guard.
Fact
The facts: on 7 August Anthropic announced an updated classifier for biology topics. By their tests, escalations to the fallback model over biology drop by about 85%. The expected drop in total escalations is 67% for Claude.ai, 55% for Cowork, 17% for Claude Code and 7% for the platform - these are projections, not measured values, and the company calls them exactly that. The fallback model is now Opus 5. Virology, toxicology and molecular design stay closed, along with professional dual-use questions and drug development. The method is described in three steps: rewriting the rules by which the classifier tells the allowed from the guarded, gathering input from internal and external experts, and retraining on new data. A separate trusted-access path for researchers is, per the text, still in development.
What pleases me here, as a builder. Until now the safety conversation ran between two extremes - either you tighten until the tool is useless, or you loosen it and hope. Anthropic does the third thing: leave the wall in place and move the door - first rewrite the rule, then check it with people, and only then touch the data.
The line isn't lowered. It's moved, and in a documented order.
Ever tried to read up on research or figure out what a metric means, and hit a wall? That's now far rarer. Don't expect a difference in coding work: the expected drop there is 17%, meaning almost nothing. And watch the 85% number - it's for biological escalations, from their own tests, not from an independent measurer.
Official primary sources
Anthropic - Improving Fable 5's Biology SafeguardsAnthropic - Claude Fable 5 and Claude Mythos 5 (the release)The visual is generated code-art, no third-party images.