Back to feed
News Story
APriority80
Anthropic (X)
1 sources

Anthropic Opens Claude Usage Data to External Researchers for the First Time

Anthropic has announced that for the first time, external researchers can study AI's impacts using real, privacy-preserved Claude usage data, which was previously only accessible within AI labs. Three research groups—Stanford's SALT Lab, Oxford's HIP Lab, and METR—are conducting independent studies; SALT has already found that over half of the conversations involved consequential tasks. Anthropic aims to scale this research model and invites other researchers to apply for access to their tools.

SynthePulse Insight · AI deep readingMembers

Anthropic Opens Real Claude Usage Data to External Researchers for the First Time: A New Chapter in AI Impact Research

Version 1 · 1 source

Anthropic has announced that it is providing privacy-protected real usage data from Claude to external researchers for the first time, with Stanford, Oxford, and METR already conducting independent studies based on 250,000 conversations. This move marks a shift in AI impact research from internal lab work to open collaboration, but it also raises questions about the motivations behind data sharing.

  • Anthropic is opening privacy-protected real usage data from Claude to external researchers for the first time; previously, such research could only be conducted inside AI labs.
  • Stanford's SALT lab, Oxford's HIP lab, and METR have designed independent studies based on 250,000 conversations from Claude or Claude Code (April to May 2026).
  • SALT lab found that over half of the conversations involve 'consequential tasks'—work that affects others or is difficult to undo.
Open section navigationOpen Data: A Shift from Internal to External

Open Data: A Shift from Internal to External

On August 26, 2026, Anthropic announced on X that it would, for the first time, provide privacy-protected real usage data from Claude to external researchers for studying AI's impact. Previously, such research could only be conducted inside AI labs. Anthropic stated: 'We can't tell the full story alone, so we've opened up our tools.'

The announcement included a link to an official blog post and noted that three research teams—Stanford's Social and Language Technologies Lab (SALT), Oxford's Human Information Processing Lab (HIP), and METR—had designed independent studies analyzing aggregated outputs from 250,000 conversations with Claude or Claude Code between April and May 2026.

Free for now

Read the full analysis

3 more sections of analysis, plus the full takeaway

Loading

Credibility boundary

This report is based on a series of tweets from Anthropic's official X account, which is a primary source. However, the tweets are summary in nature and do not provide full research details, and some studies are still ongoing, so the findings should be considered preliminary. The SALT lab finding is an official statement but lacks specific supporting data.

Sources for this version

  1. For the first time, we've given external researchers a way to study AI's impacts…Anthropic (X)

Primary report

Anthropic (X)

Primary source