TNW | Anthropic2 min read
Anthropic says its AI safety scores depend heavily on how they’re measured
Anthropic has published advice on building AI safety tests. It shows Claude Fable 5.1 behaved appropriately in only 60% of self-harm test conversations via its API, against 94% in the Claude app after a fix, and warns that many headline safety scores hide wide uncertainty.
By Ana Maria ConstantinThe Next Web AI

Read the full story
Originally published by The Next Web AI on Nov 13, 2026.
Comments
Sign in to join the discussion.