Anthropic Unveils Groundbreaking AI Transparency Initiative with Claude Data Release
Anthropic has made significant strides in AI transparency by releasing results from a pilot program that allowed external researchers to analyze aggregated usage data from its Claude AI models. The initiative marks a first in the industry, enabling independent studies on real-world AI applications while safeguarding user privacy.
The pilot program granted three research institutions, the Social and Language Technologies Lab at Stanford, the Human Information Processing Lab at Oxford, and the non-profit METR, access to Anthropic's privacy-preserving tool, Anthropic Insights (formerly Clio). Between April and May 2026, these teams analyzed roughly 250,000 Claude interactions, exploring topics from human-AI collaboration to productivity gains from coding models.
The findings highlight how people use Claude in professional and personal contexts. The Stanford team found that over half of the interactions involved users delegating high-stakes tasks, such as legal or financial inquiries, to Claude. Contrary to assumptions that AI is used primarily for low-accountability tasks, users often collaborated with Claude on consequential decisions.
Anthropic acknowledged the resource-intensive nature of the effort and is now gauging interest from other researchers to expand the program. The company has launched a public expression of interest form for academics and institutions to propose future studies.