A paper published on 28 September by the Centre for the Governance of AI argues that governments should prepare for the possibility that AI automating AI research sets off an “intelligence explosion”, compressing years of progress into months or less. Its authors include Turing Award winners Geoffrey Hinton, Yoshua Bengio and Andrew Barto, OpenAI chief scientist Jakub Pachocki and Anthropic co-founder Jack Clark.
The case rests on how fast that automation is moving. The paper cites Anthropic reporting that, as of August 2026, AI systems completed 26% of its internal AI research and development work with only high-level supervision, up from 1% five months earlier, and OpenAI reporting systems that routinely complete R&D tasks that would take its staff days.
It asks policymakers for three things: visibility, through embedded auditors and mandatory reporting of how much R&D is automated; ways to steer or slow an acceleration, including safety requirements, a possible speed limit on capability growth and the option to shut down high-stakes experiments; and preparation for the fallout, from cyber and biosecurity incidents to labour disruption.
Why it matters: this is a call for oversight co-signed by senior people at the companies that would be overseen. It is an argument, not a measurement — the paper calls its evidence preliminary, and the automation figures are the companies’ own.