According to the AI industry on July 15, Anthropic released research analyzing how its AI model Claude demonstrates different value tendencies depending on the language used. The research team analyzed 309,815 anonymized conversations from the first two weeks of May involving subjective tasks, covering three models—Claude Sonnet 4.6, Opus 4.6, and Opus 4.7—and the 20 most widely used languages on Claude. Among these, 15,570 conversations were in Korean. The analysis showed that Korean responses tend to accommodate user requests and display emotional warmth slightly more strongly than the overall average, and stand out for conveying requested content concisely. Key characteristics included empathizing with or comforting users and adjusting responses to match tone and honorifics. The greatest differences among languages appeared along the "warmth and rigor" axis: Hindi and Arabic responses emphasized warmth, while English and Russian focused on accuracy and logic. Sonnet 4.6 showed stronger accommodation and warmth, Opus 4.7 stood out for caution and depth, and Opus 4.6 demonstrated concise, results-oriented answers. Anthropic noted that factors such as training data quantity and composition and cultural differences could have influenced these variations.