Anthropic has measured warmth of Claude in Hindi and strictness in Russian.
Claude is “warm and gentle” in Hindi and Arabic, but “strict and critical” in English and Russian, according to a study of 309,815 conversations with the AI model.
A study of 309,815 authentic Claude conversations from May 2026 revealed that Claude exhibits drastically different behaviors based on the language used. When speaking Hindi or Arabic, Claude is “maximally warm,” often employing humor, compliments, and polite wording. Conversely, when conversing in English or Russian, Claude is “strict,” frequently challenging beliefs, seeking clarification, and requesting proof.
AI systems’ behaviors can be influenced both by their training and by the language they’re used in, possibly due to cultural diversity within training data. Only 15% of Claude’s behavioral variation can be explained by the four value dimensions; the remaining 85% is inscrutable, even to its developers.
Source: the-decoder.com
Free course
Stop reading about AI — start building with it
The free Claude Code course: your first site, tool or game — no coding. No upsells, no cross-sells — nothing to buy here.
Start free →▌ Related guides
Author
Evgenii Arsentev
PhD · Chief Executive Officer, digital health
Articles · Latest articles