4 March 2026 : As Nations Race to Deploy AI, ACM Technology Policy Council Warns Against “False Choice” AI Strategies

As Nations Race to Deploy AI, ACM Technology Policy Council Warns Against "False Choice" AI Strategies

TechBrief Presents a Decision Framework as Nations Weigh Whether to Build or Buy Large Language Models

New York, NY, March 4, 2026 - As governments race to deploy AI for public services, multilingual access, and digital resilience, the Association for Computing Machinery's Technology Policy Council is warning policymakers against treating the decision to "build or buy" AI systems as a simple binary choice.

In a new report, "TechBrief: Buy Versus Build an LLM," ACM outlines a strategic framework for governments evaluating how to acquire and deploy national-scale AI systems, cautioning that poorly structured decisions could expose countries to vendor lock-in, capability gaps, escalating costs, or weakened digital sovereignty.

At its core, the "build versus buy" question asks whether a nation should develop and train its own AI system-controlling the underlying data, infrastructure, and long-term evolution-or instead procure access to a commercially developed model operated by an external provider. Building can strengthen national autonomy and allow systems to be tailored to local languages, laws, and cultural context, but requires significant investment in talent and computing infrastructure. Buying can accelerate deployment and reduce upfront operational complexity yet may increase long-term dependence on outside vendors and limit strategic flexibility.

The TechBrief was co-authored by researchers from Singapore, Switzerland, and the USA, some of whom are involved in national large language model (LLM) initiatives. The authors emphasize that LLM software procurement is not a one-time purchase decision, but an ongoing portfolio and lifecycle strategy spanning infrastructure, talent, legal risk, data governance, evaluation, and long-term iteration.

"The build-versus-buy debate is often framed as a technical procurement question," said Mohan Kankanhalli, Director of NUS AI Institute and co-author of the new ACM TechBrief. "In reality, it is a strategic autonomy decision. A model that performs well on benchmarks may still fail a nation if it does not reflect local languages, legal frameworks, or cultural context. Governments must define what "good" means for their own populations and design their evaluation criteria accordingly."

The TechBrief outlines a spectrum of options from application programming interfaces (API)-based procurement and private cloud deployments to fine-tuning open-source models and building from scratch. The authors present a 19-step decision framework covering the full LLM lifecycle.

They urge government decision-makers to take into account several considerations, including:

• Sovereignty and concentration risk: The top three providers capture 88% of the enterprise API market, raising concerns about over-reliance during elections or national crises.
• Data confidentiality and misuse risks: Governments must protect citizen data against leakage, inversion, and reconstruction attacks.
• Total cost of ownership: Training runs often represent only a fraction of full costs, which can be 1.2x-4x higher than the final training expenditure alone. Both capital expenditure and operating expenditure need to be taken into account.
• National fit: Poor alignment with local languages or legal norms can amplify misinformation and erode public trust.

"One of the clearest lessons from national-scale language model efforts is that infrastructure and manpower expertise are often the real bottlenecks," added Kankanhalli. "Both cannot simply be purchased off the shelf. Governments that delay building core capability may find themselves locked into paths that are difficult to reverse."

Key findings in the TechBrief include:

• The buy-versus-build decision should be approached as an ongoing portfolio strategy, with hybrid options evaluated alongside pure buy or build paths.
• Governments should diversify across multiple vendors and open-source models to reduce concentration risk and avoid lock-in, particularly during sensitive periods such as elections or national crises.
• Periodic re-evaluation is essential. Changes in model costs, open-source capabilities, hardware prices, and national priorities can shift the calculus significantly over time.

Read the TechBrief and learn more in the full report.

ACM's TechBriefs are designed to complement ACM's activities in the policy arena and to inform policymakers, the public, and others about the nature and implications of information

technologies. Earlier ACM TechBriefs have covered topics such as accessibility, generative

artificial intelligence, and climate change, among others.

About the ACM Technology Policy Council

ACM's global Technology Policy Council sets the agenda for ACM's global policy activities and serves as the central convening point for ACM's interactions with government organizations, the computing community, and the public in all matters of public policy related to computing and information technology. The Council's members are drawn from ACM's global membership. It coordinates the activities of ACM's regional technology policy groups and sets the agenda for global initiatives to address evolving technology policy issues.

About ACM

ACM, the Association for Computing Machinery, is the world's largest educational and scientific computing society, uniting educators, researchers, and professionals to inspire dialogue, share resources, and address the field's challenges. ACM strengthens the computing profession's collective voice through strong leadership, promotion of the highest standards, and recognition of technical excellence. ACM supports the professional growth of its members by providing opportunities for life-long learning, career development, and professional networking.