Claude constitution's 2023 vs 2026 wording: from rule lists to a empathy heuristic
teortaxesTex · x · 2026-11-18
A side-by-side comparison of Anthropic's Claude constitution shows a striking shift: the 2023 version used direct prohibitions (don't choose toxic, racist, or unethical responses), while the January 2026 version replaces them with a judgment heuristic — when unsure if it's being overcautious or overcompliant, Claude should imagine how a thoughtful senior Anthropic employee who deeply cares about doing the right thing might react. Commenters note the new phrasing feels very much like Dario's sensibility: hard rules giving way to internalized organizational judgment.
More from AGI Musings
- Investor claims media is fake fear-mongering AI outages to stoke hate — econoar · 2026-11-18
- We Are the Misaligned AI: an ML Practitioner's Evolution Analogy for Alignment — theteknosaur · 2026-11-18
- RLHF inventor Paul Christiano: 1-in-3 chance of uncontrollable superintelligence within 12 months — 新智元 · 2026-11-18
- Oxford researcher's 59-page paper proposes the 'Cognitive Symbiont' and AHI as AI's other endgame — 新智元 · 2026-11-18
- Noah Smith predicts 4.6% annualized per-capita US real GDP growth through 2060 — inductionheads · 2026-11-18
- Timothée Chalamet is reading Kevin Roose's 'The AGI Chronicles' — and lent it to Josh Safdie — kevinroose · 2026-11-18