This report assesses public procurement as an AI governance tool, develops contractual guarantees for state buyers, and examines whether coordination among AI middle powers could strengthen bargaining leverage, evaluation capacity and resilience to AI crises.

Public procurement and AI governance

Developing a strategic framework for middle powers

Afek Shamir

ResearchPublished Sep 11, 2026

Frontier AI capabilities are concentrated among a small number of private firms headquartered in the United States and China. For states which neither develop nor host leading models, this concentration creates a dependence on foreign AI providers whose capabilities they often cannot independently evaluate and whose technology may be constrained by the laws of the provider’s home jurisdiction.

Public procurement offers ‘AI middle powers’ one mechanism for governing these dependencies. These are states with significant economic, diplomatic, security or regulatory capabilities that nevertheless depend technologically on foreign frontier AI firms. The report finds that three categories of risk fall disproportionately on state buyers in AI middle powers: vendor lock-in and the threat of service withdrawal, limited visibility into frontier AI development, and data sovereignty and information-security risks.

To mitigate these risks, the report develops a framework of procurement obligations that governments could seek to negotiate with AI providers. These include model access for evaluation, information sharing, funding for reliability research, state-of-the-art risk management requirements, termination and exit provisions, and measures to support domestic and allied capability.

The analysis finds that public procurement is a useful but bounded mechanism for governing frontier AI. Governments are not yet major customers of frontier AI firms, and leading developers may be able to forego government contracts rather than accept costly requirements. Procurement contracts with foreign providers may also remain subordinate to the national-security laws of a vendor’s home state, and state buyers can face persistent information asymmetries when negotiating with leading technology companies. At the same time, government demand for frontier AI is growing, and public-sector contracts offer firms more than revenue alone. They provide financial predictability and a degree of prestige and regulatory validation, making governments strategically important customers even if they are not yet indispensable to firms’ business models.

The central recommendation is that AI middle powers should assess whether collective coordination on procurement requirements can strengthen their bargaining position and state capacity.

Key Takeaways

Short-term actions (actions that could be initiated within 12 months):

  • Firstly, governments should begin with a shared assessment of the risks that procurement requirements are intended to address. Participating states should compare the risks they associate with different government uses of frontier AI, including but not limited to dependence on individual foreign providers, loss of access, insufficient visibility into model behaviour, and requirements for data control. Common procurement provisions should follow predominantly where separate governments assess risks similarly.
  • Secondly, governments should assess the most suitable institutional arrangement for procurement coordination. The G7, the EU, NATO, and International Network of AI Safety Institutes already provide mechanisms for cooperation on AI policy, standards and technical evaluation. A dedicated arrangement may be justified where no current institution combines sufficient purchasing relevance and technical capacity.
  • Thirdly, smaller middle powers which lack state access to frontier AI should explore opportunities to pool demand with partners and strengthen their negotiating position. Where an individual government is unlikely to represent sufficient demand to influence a frontier AI provider, cooperation with other buyers could increase the commercial significance of their purchasing preferences and requirements.

Medium- to long-term actions (actions likely to require approximately 1–5 years or longer):

  • Firstly, governments should identify which AI capabilities need to remain available independently of any single foreign provider and invest selectively in the infrastructure required to preserve them. For most middle powers, this is unlikely to mean reproducing the full frontier AI stack nationally. A more realistic strategy could include building a meaningful share of domestic or allied compute, creating secure hosting facilities, securing access to multiple model families for optionality reasons, and bringing in technical talent to government with expertise in fine-tuning and the know-how to migrate critical workloads. These assets would also strengthen bargaining power by making the future option of switching providers more credible.
  • Secondly, governments should strengthen their capacity to evaluate and assure frontier AI systems. This includes developing assurance and technical AI expertise within defence agencies and supporting the creation of a mature third-party evaluation and assurance sector. These capabilities are necessary to make procurement guarantees practically enforceable.
  • Thirdly, procurement coordination could provide a basis for wider cooperation on AI incidents and resilience to AI-related failures. Governments that establish common procurement requirements would have an existing mechanism for sharing information with one another and with providers. Over time, this could support common approaches to incident reporting and exercises for responding to significant AI-related incidents.

Topics

Document Details

Citation

Chicago Manual of Style

Shamir, Afek, Public procurement and AI governance: Developing a strategic framework for middle powers. Santa Monica, CA: RAND Corporation, 2026. https://www.rand.org/pubs/research_reports/RRA5284-1.html.
BibTeX RIS

Research conducted by

This research was conducted within the Science and Emerging Technology Program within RAND Europe.

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.