Fable 5's Cyber Safeguards and Jailbreak Severity Framework
Source
anthropic.com
Date
Why it matters
If you build security tooling on Fable 5, this states where the cyber classifiers draw the line and proposes a severity scale for grading jailbreakA prompt crafted to make a model ignore its own guidelines — usually through roleplay, hypotheticals, or encoding rather than a direct request.Full definition → attempts rather than treating every bypass as equivalent.
Terms in this piece · Glossary
jailbreak — A prompt crafted to make a model ignore its own guidelines — usually through roleplay, hypotheticals, or encoding rather than a direct request.
