Claude responds with more warmth in Hindi and more rigor in Russian, showing how language shapes AI answers

Anthropic has released a new study examining the values expressed by Claude models, mapping hundreds of value concepts derived from thousands of terms. The study highlights systematic differences across Claude models and languages, showing that the model responds with more warmth in Hindi and more rigor in Russian. These findings raise important questions about how language influences AI responses, suggesting a need for further research in this critical area.