x.com/sporadica/status/2064400134700695830
1 correction found
if you ask Fable (Mythos) anything related to cybersecurity or bio, it's gonna redirect you to Opus 4.8
Anthropic does not say all cyber- or bio-related prompts are rerouted. Its own docs say only a narrow set of safeguarded topics trigger fallback to Opus 4.8, mainly offensive cyber requests and certain biology/life-sciences content, though some benign cases may also be caught.
Full reasoning
Anthropic's own announcement on X says "Queries on a narrow range of topics" are answered by Opus 4.8, not anything related to cybersecurity or biology.
Its API documentation is more specific: Claude Fable 5 runs safety classifiers targeting offensive cybersecurity techniques such as exploit-building, malware, or attack tooling, plus certain biology and life sciences content such as lab methods or molecular mechanisms. Anthropic also notes that some benign cybersecurity or beneficial life-sciences tasks may also trigger these safeguards.
So the post overstates the policy. The official description is not "anything related to cybersecurity or bio"; it is a narrower safeguard with some false positives possible.
2 sources
- Claude on X: "Releasing a model this capable comes with risks..."
Queries on a narrow range of topics will instead receive a response from our next-most-capable model, Opus 4.8.
- Prompting Claude Fable 5 - Claude API Docs
Claude Fable 5 runs safety classifiers that target offensive cybersecurity techniques (such as building exploits, malware, or attack tooling), biology and life sciences content (such as lab methods or molecular mechanisms)... Benign cybersecurity work and beneficial life sciences tasks may also trigger these safeguards. To re-route declined requests automatically, configure server-side or client-side fallback to Claude Opus 4.8.