Stop The
AI Race

China would also need to pause. Any agreement only works if it includes all major AI labs globally. The MIRI Technical Governance Team published a detailed proposal for an international agreement centered on a coalition led by the US and China, with verification mechanisms including AI chip tracking. Public commitments from Western lab CEOs are a first step toward the kind of international coordination that makes this possible.

No new training runs of larger or more general frontier models. The teams currently working on improving the capabilities of these models would move to narrow AI applications or alignment research instead. Current models stay available. Narrow AI applications are fine. The principle is more important than the specific definition of "frontier," but for a detailed technical proposal, see MIRI's international agreement paper, which operationalizes limits through compute thresholds (FLOP caps) and restricts research that advances toward superintelligence while preserving access to current AI.