Alan Chan, Sören Mindermann et al. (22 authors) · 2026-09-28

What if automating AI R&D triggers an intelligence explosion?

“AI systems are on track to automate most AI R&D work within a few years, and possibly all of it.”

Who: Alan Chan, Sören Mindermann et al. (22 authors)
Date: 2026-09-28 First posted on arXiv on 2026-09-28. The text we archived is version 2, posted 2026-10-26.
Evidence: Not rated. A policy paper built on other studies' evidence. It reports no new measurement of its own. how we rate evidence
Replication: Nothing to rerun. all claims
Checked: 2026-10-28, against the source page
Links: Source Paper

Also on the page

The authors include Jakub Pachocki (OpenAI), Jack Clark (Anthropic), Geoffrey Hinton and Yoshua Bengio. The paper says the views are the authors' and "do not necessarily represent the views of the organizations with which they are affiliated."
On timing, it cites tentative extrapolations of recent trends: "months-long AI R&D projects will be automated by mid-2028".
It gives Anthropic's figure as rising "from 1% to 26% between March and August 2026". Anthropic's own page says 26% as of August 2026, "up from under 1% in February 2026".
Its policy asks: "Policymakers should urgently obtain more visibility into the automation of AI R&D, develop ways to steer and constrain an intelligence explosion, and prepare society to adapt to an intelligence explosion’s impacts."