Anthropic says these topics are too dangerous to let its Fable 5 model talk about
Summary
Anthropic released Claude Fable 5, a new AI model with stronger safety controls that block it from discussing sensitive topics like cybersecurity, biology, and chemistry. The model performs better than earlier versions but redirects risky queries to a previous model and warns users to prevent misuse.Key Facts
- Claude Fable 5 is Anthropic’s first "Mythos-class" AI model, improving on previous Claude Opus models.
- It restricts answering questions on cybersecurity, biology, chemistry, and other sensitive areas to prevent harm.
- Fable 5 redirects some restricted queries to the older Claude Opus 4.8 model and notifies users when this happens.
- Anthropic made the safeguards stricter, which may sometimes block safe questions but helps reduce misuse risks.
- Over 1,000 hours of testing showed no universal jailbreaks to bypass Fable 5’s safeguards.
- The model shows a large improvement in cybersecurity tasks, scoring 78% on a vulnerability benchmark, up from 40% in older models.
- Anthropic is particularly concerned about cyberattacks automated by AI and risks from biological research queries.
- Access to the more powerful Mythos 5 model is still limited to trusted cybersecurity experts through a special program.
Read the Full Article
This is a fact-based summary from The Actual News. Click below to read the complete story directly from the original source.