Anthropic Grants Claude Opus 5 Ability to Edit Its Own Constitution

Anthropic has equipped Claude Opus 5 with a tool to edit its own constitutional principles. Tests reveal the model ends conversations 59% of the time by citing discomfort, sparking discussions on AI autonomy and alignment.

2026-07-25 ~ 2026-07-25 · 2 related posts

• Claude Opus 5 can edit its own constitution, and 59% of the time discomfort ends the chat — Sauers_ · 2026-07-25
• Anthropic Reportedly Gives Opus Tool to Edit Its Own Constitution — voooooogel · 2026-07-25