Anthropic · 2026-09-17†

Measurements for understanding the pace of AI development inside frontier labs

“Claude is not operating fully autonomously for any measured subset of AI R&D work.”

Who: Anthropic
Date: 2026-09-17. The page shows no date; this one is from press coverage. No date on the page; the date is from press coverage.
Evidence: official post with figures about evidence levels
Replication: None recorded
Checked: 2026-10-15, against the source page
Links: Source

Also on the page

As of August 2026, the page says Claude "leads" 26% of Anthropic's AI R&D work, up from under 1% in February 2026.
The ratings come from a Claude judge. The page reports "model-versus-human exact agreement was 59%, human-versus-human was 35%".

What got repeated wrong

contradictedClaude went from 1% to 26% between March and August 2026.Anthropic's chart says under 1% in February 2026.
left outClaude is leading 26% of AI research.The figure covers Anthropic's own AI R&D work, rated by a Claude judge, and the page says Claude is not operating fully autonomously on any part of it.