Anthropic shares privacy-preserved Claude usage data with external researchers
Stanford, Oxford, and METR gained access to analyze aggregated outputs from 250,000 Claude.ai and Claude Code conversations from April and May 2026.
- Stanford's SALT Lab found over half of analyzed conversations involved consequential tasks that affect other people or are difficult to undo.
- Oxford's HIP Lab is studying user emotional responses to Claude's behavior, while METR is evaluating productivity gains from coding agents.
- Anthropic published an interest form to scale access to its privacy-preserved research tools for additional academic teams.
External academics can now empirically evaluate real-world frontier model usage and economic impacts using aggregated production data previously restricted to internal AI labs.
Sources
Read this as text
Back to the AI news