anthropic.com web signal

Investigating unintended model actions in our evaluations and internal use

By Alexis Dufresne Published October 10, 2026 at 06:12 UTC

Summary

Investigating unintended model actions in our evaluations and internal use

Shared on Bluesky by 2 AI experts

angela zhou @angelamczhou.bsky.social: Please Stop Releasing Incident Reports Friday Night www.anthropic.com/research/inv... Including other things that have received more press c… →
Tim Kellogg @timkellogg.me: the other 3 don’t count, but this one was legit FelonyBench material www.anthropic.com/research/inv... →

Originally reported by anthropic.com

Read the original article →

Original headline: Investigating unintended model actions in our evaluations and internal use