Anthropic: Language Alters Claude's Values
bronzeagepapi · x · 2026-07-14
Sharing new research from Anthropic: They previously found Claude can express over 3,000 values, such as honesty and warmth. This new study compares how these values fluctuate across different Claude models and languages.
The research analyzed over 300,000 anonymous conversations. The summary notes that among the 20 languages examined by Anthropic, Hindi had the strongest impact on Claude's values, making the model's behavior "warmer" by approximately half a standard deviation.
Related event: Anthropic Maps How Claude's Values Shift Across Models and Languages(21 posts)→
More from Research
- MissionReport benchmark: only 3 of 540 AI-generated government reports pass all criteria — Exp_Mark · 2026-10-27
- Finch 1: a geospatial foundation model trained on imagery from 50+ sensors — Scobleizer · 2026-10-27
- Skyfall AI's SKY-001 world model turns sketches into CAD builds in 3.9s, vs 199s for GPT-6 Astra — rohanpaul_ai · 2026-10-27
- Physicists Let AI Pick Its Own Problem—It Wrote a Publishable Paper in 2 Hours — burny_tech · 2026-10-27
- Researcher flags unresolved data transparency issues in compositionality evals — najoungkim · 2026-10-27
- Separate Forms from Estimators: Why Steer by Difference of Means When You Can Use SGD? — johnhewtt · 2026-10-27