This report documents research and analysis conducted as part of a study to investigate detection and monitoring mechanisms that cloud service providers could employ to identify large artificial intelligence (AI) training runs. The intended audience of this report is national policymakers interested in compute-based AI governance. This report may be of interest to cloud service providers and infrastructure-as-a-service companies.

Strategies and Detection Gaps in a Game-Theoretic Model of Compute Governance

Alvin Moon, Padmaja Vedula, Jesse Geneson, Simon Bar-on

ResearchPublished Jun 16, 2025

This report documents research and analysis conducted as part of a study to investigate detection and monitoring mechanisms that cloud service providers could employ to identify large artificial intelligence (AI) training runs. The intended audience of this report is national policymakers interested in compute-based AI governance. This report may be of interest to cloud service providers and infrastructure-as-a-service companies.

Key Findings

  • Cloud service providers might not be able to report and detect AI training if they are only obligated to monitor and report activity based on floating-point operation thresholds in future AI governance policy.
  • In the future, many types of capable AI models will likely exist whose training will be hard to detect. The authors outline strategies for cloud service providers that could aid with AI training detection in cloud computing environments.

Recommendations

  • To develop effective AI governance, policymakers should support efforts to find detection gaps in compute-based monitoring schemes.
  • Policymakers should continue to pursue both compute- and noncompute-based AI governance.
  • Continuing research into effective thresholds for compute monitoring is required to create a robust compute-based monitoring framework that can adapt to technological progress.

Topics

Document Details

Citation

Chicago Manual of Style

Moon, Alvin, Padmaja Vedula, Jesse Geneson, and Simon Bar-on, Strategies and Detection Gaps in a Game-Theoretic Model of Compute Governance. Santa Monica, CA: RAND Corporation, 2025. https://www.rand.org/pubs/research_reports/RRA3686-1.html.
BibTeX RIS

Research conducted by

This research was independently initiated and conducted within the Technology and Security Policy Center of RAND Global Emerging Risks using income from operations and gifts from philanthropic supporters. A complete list of donors and funders is available at www.rand.org/TASP.

This publication is part of the RAND research report series. Research reports present research findings and objective analysis that address the challenges facing the public and private sectors. All RAND research reports undergo rigorous peer review to ensure high standards for research quality and objectivity.

This document and trademark(s) contained herein are protected by law. This representation of RAND intellectual property is provided for noncommercial use only. Unauthorized posting of this publication online is prohibited; linking directly to this product page is encouraged. Permission is required from RAND to reproduce, or reuse in another form, any of its research documents for commercial purposes. For information on reprint and reuse permissions, please visit www.rand.org/pubs/permissions.

RAND is a nonprofit institution that helps improve policy and decisionmaking through research and analysis. RAND's publications do not necessarily reflect the opinions of its research clients and sponsors.