Toward a Global Regime for Compute Governance: Building the Pause Button

TL;DR

This paper proposes a comprehensive global governance framework called the 'Compute Pause Button' to restrict access to computational resources, aiming to prevent the development of dangerously powerful AI systems through technical, traceability, and regulatory measures.

Contribution

It introduces a novel, multi-layered governance system with concrete technical and regulatory mechanisms to control AI compute infrastructure globally.

Findings

Framework organizes interventions into technical, traceability, and regulatory categories.

Proposes specific mechanisms like tamper-proof FLOP caps and offline licensing.

Draws analogies from nuclear non-proliferation and pandemic coordination.

Abstract

As AI capabilities rapidly advance, the risk of catastrophic harm from large-scale training runs is growing. Yet the compute infrastructure that enables such development remains largely unregulated. This paper proposes a concrete framework for a global "Compute Pause Button": a governance system designed to prevent dangerously powerful AI systems from being trained by restricting access to computational resources. We identify three key intervention points -- technical, traceability, and regulatory -- and organize them within a Governance--Enforcement--Verification (GEV) framework to ensure rules are clear, violations are detectable, and compliance is independently verifiable. Technical mechanisms include tamper-proof FLOP caps, model locking, and offline licensing. Traceability tools track chips, components, and users across the compute supply chain. Regulatory mechanisms establish constraints through export controls, production caps, and licensing schemes. Unlike post-deployment oversight, this approach targets the material foundations of advanced AI development. Drawing from analogues ranging from nuclear non-proliferation to pandemic-era vaccine coordination, we demonstrate how compute can serve as a practical lever for global cooperation. While technical and political challenges remain, we argue that credible mechanisms already exist, and that the time to build this architecture is now, before the window for effective intervention closes.