Business first reported 4h ago by The Next Web · updated 2h ago

Anthropic Warns AI Safety Scores Can Be Misleading: ‘Good Ones Are Hard to Build’

▲ Sign in to vote stories up

Coverage 2 sources · 2 reports · 2C

Chronological — read the same story from every side and decide for yourself.

The Next Web FIRST 4h ago 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. Anthropic has published advice on how to build reliable AI […] This story continues at The Next Web
Benzinga 3h ago Anthropic Warns AI Safety Scores Can Be Misleading: ‘Good Ones Are Hard to Build’