An International AI Slowdown Is Ready Whenever Politicians Are

Per LessWrong, more than a thousand employees from the world's top AI companies have called for an international agreement to slow artificial intelligence development, with both OpenAI and Anthropic now officially backing the effort. Skeptics argue such a slowdown would be impossible to enforce—how could the US verify that China is actually slowing down? But LessWrong proposes a solution: deploying human inspectors directly inside AI labs, with both physical access to hardware and digital access to code, logs, and communications. This "whole lab inspection" approach could enforce restrictions on dangerous capabilities—preventing AI systems from autonomously designing smarter AI systems, and limiting how models are improved after training. The technical infrastructure already exists; the question is whether governments will muster the political will to implement it.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton