Discover · thenextweb.com

Anthropic says its AI safety scores depend heavily on how they’re measured

The Next Web thenextweb.com

This is what was read here when it was published. Ana Maria Constantin no longer carries it in what it publishes now.

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…

More ways to use this
Open on another device
Scan this QR code to open the same public page

Scan to open this page. Your saved notes and account access are not included.

Download QR code

Open the page link

Names and references

  1. Anthropic — person

Keep going

Each of these is read from what this estate already kept: the feed a piece came through, the publisher behind it, the credits and topics the publisher printed. None of it is a recommendation, an identity or a record kept anywhere else.

Also from The Next Web

  1. LangChain raises $300m at a $3bn valuation to help companies run AI agentsListed at $300 Cristian Dina thenextweb.com
  2. Cognition and McKinsey team up to sell Devin, its AI coder, to big firmsCristian Dina thenextweb.com
  3. Databricks pitches AI that predicts from tables without trainingCristian Dina thenextweb.com
  4. Google’s AdMob lets AI agents like Claude Code set up app adsAna Maria Constantin thenextweb.com
  5. OpenAI lets Apollo embed six AI safety testers during trainingAna Maria Constantin thenextweb.com
  6. Stanford professors’ AI lab Noeri launches with $110m, backed by NvidiaListed at $110 Cristian Dina thenextweb.com