AIThe Registerabout 2 hours ago
Anthropic study maps Claude values across languages
4 min read
Anthropic analyzed 309,815 anonymized conversations from May 2026 to map value shifts in Claude models across 20 languages. Four core axes emerged: Deference and Caution, Warmth and Rigor, Depth and Brevity, and Candor and Execution. Sonnet 4.6 shows more warmth and deference; Opus 4.7 warns about risks unprompted. Hindi and Arabic responses show the most warmth; English and Russian show more rigor.
Level
Hype check
Tap to vote and see what everyone thinks.
In this storyAnthropic
