AI Slowdown: Who Would Actually Enforce It?
Original: Here’s How an AI Slowdown Could Actually Be Enforced
Why This Matters
RSI risk is pushing AI governance from abstract debate toward concrete enforcement design—a field that barely exists yet.
As AI lab leaders including Dario Amodei, Sam Altman, and Demis Hassabis voice support for an AI pause, researchers warn that enforcing any slowdown remains an unsolved technical and governance problem. A new University of Toronto report, 'Pacing the Frontier,' calls for dedicated research into what enforcement would even look like.
The call for an AI slowdown is getting louder—but nobody has a clear plan for making one stick. A new report from University of Toronto researcher Raymond Douglas, titled 'Pacing the Frontier: A Research Agenda,' lays out just how thin the current thinking is. 'We don't really understand what our options even are or what they will do,' Douglas says.
The urgency has spiked after a departing Anthropic researcher warned that AI could threaten humanity within a couple of years—a view the head of Anthropic's own safety lab publicly backed. CEOs across the industry, including Sam Altman (OpenAI), Elon Musk (SpaceXAI), and Demis Hassabis (Google DeepMind), have now echoed support for some form of slowdown.
The specific fear driving much of this is recursive self-improvement (RSI): AI systems now help build future AI systems, potentially creating an accelerating loop that outpaces human oversight. Anthropic this week published new tracking metrics showing Claude now handles 26% of the company's internal AI research—up from zero at the start of 2026—and that 6% of its compute budget goes toward safety work.
Proposed enforcement mechanisms range from third-party model audits to GPU tracking devices to chip destruction ceremonies. Geoffrey Irving, former chief scientist at the UK AI Security Institute, argues that rigorous inspections could work in the near term: 'Inspections and audits work, or even just mutual agreements.' But experts broadly agree that any credible enforcement system will need independent funding and expertise well outside the labs themselves.