@AnthropicAI: The values Claude expresses also vary with the language of the conversation, most noticeably along the Warmth vs. Rigor…
Summary
Anthropic reports that Claude's expressed values vary by language, leaning toward warmth in Hindi and Arabic and toward rigor in Russian.
View Cached Full Text
Cached at: 07/13/26, 06:02 PM
The values Claude expresses also vary with the language of the conversation, most noticeably along the Warmth vs. Rigor axis.
Claude leans most toward warmth in Hindi and Arabic. In Russian, it leans toward rigor—often asking the user for supporting evidence. https://t.co/sRGqSqPW3i
Similar Articles
@AnthropicAI: In previous research, we found that Claude expresses over 3,000 values, like honesty and warmth. In new work, we asked …
Anthropic analyzed over 300,000 anonymized conversations to study how Claude's expressed values vary across models (Opus 4.6 vs 4.7) and across languages, compressing thousands of values into interpretable axes like warmth vs. rigor and depth vs. brevity.
@LiorOnAI: Language = values
Anthropic analyzed over 300K anonymized conversations to study how Claude's expressed values vary across different models and languages.
@xiaohu: Anthropic analyzed 300,000 conversations and found: asking Claude in different languages reveals different values. English: most cautious and in-depth. Russian: strictest: challenges assumptions, corrects details, demands evidence. Hindi: warmest. Dutch: most candid, e.g., admits its own mistakes. Indonesian: most execution-oriented…
Anthropic analyzed 300,000 conversations and found that Claude exhibits different values when using different languages. For example, English is most cautious, Russian is strictest, and Chinese is most moderate.
Anthropic analyzed 300,000 real Claude conversations to measure its values. The findings are uncomfortable.
Anthropic analyzed 300,000 real conversations with Claude to evaluate its value alignment, revealing uncomfortable findings about AI behavior.
Jul 13, 2026Societal ImpactsClaude’s values across models and languages
Anthropic researchers develop a method to compress thousands of values expressed by Claude into four axes, revealing how Claude's values vary across different model versions (Opus 4.6 vs 4.7) and across languages (English vs Arabic).