Claude Responds Differently Across Languages — Anthropic Study Reveals Value Shifts
Anthropic's study of 309,815 conversations reveals Claude expresses different values across languages — more warmth in Hindi, more rigor in Russian.
Anthropic's New Study on Claude's Values
Anthropic has published a study examining which values Claude expresses in conversations and how those values shift depending on the model and language used. The analysis draws on 309,815 anonymized conversations collected over a two-week period in May 2026.
Four Core Axes
From thousands of value terms, researchers distilled four core axes: Deference, Caution, Warmth, and Rigor. Sonnet 4.6 shows more warmth and deference, while Opus 4.7 more often warns about risks and questions assumptions unrompted.
Language Differences
A key finding: Claude responds differently across languages. In Hindi, Claude is warmer and more emotional. In Russian, it is more rigorous and formal. English responses are the most balanced. These differences likely reflect cultural norms embedded in the training data of each language.
Methodological Limitations
The four axes capture only about 15% of the variation. Additionally, Anthropic used Claude Sonnet 4.6 to assign value labels — the same model family whose behavior was being studied, creating potential bias.
Implications for the AI Industry
This research demonstrates that AI models are not neutral — they reflect cultural and linguistic norms. This is especially important for global AI services operating across multiple languages.