UNDP Calls for Stronger AI Governance as Public Services Face Rising Trust and Safety Challenges

A UNDP report finds that AI is spreading through government systems faster than institutions can govern it, making stronger procurement rules, digital infrastructure, accountability, and trust mechanisms essential for safe adoption. It urges governments, development partners, and the private sector to prioritize practical AI governance and regional cooperation to ensure AI drives inclusive and sustainable development rather than increasing institutional risks.

Artificial intelligence (AI) is no longer confined to research labs or technology companies. It is quietly becoming part of government offices, hospitals, schools, banks, courts, and welfare systems around the world. However, a new United Nations Development Programme (UNDP) report warns that while AI adoption is accelerating rapidly, the institutions responsible for managing its risks are struggling to keep pace. The report, based on engagements across 11 countries and territories between November 2025 and March 2026, argues that the world's biggest AI challenge is no longer building smarter models but ensuring governments can deploy AI safely, transparently, and responsibly. The findings have significant implications for policymakers, international development organizations, and private-sector technology providers because the decisions made today will determine whether AI becomes a driver of inclusive development or a source of new governance risks.

AI Is Spreading Faster Than Governments Can Govern It

The report finds that AI is increasingly entering public institutions through software updates, cloud-based enterprise platforms, and everyday workplace tools rather than through carefully planned national AI strategies. Civil servants often use AI applications for drafting documents, legal research, recruitment, and administrative work without formal approval or oversight. This growing phenomenon, sometimes called "shadow AI," makes it difficult for governments to know where AI is being used, how decisions are influenced, or what risks are emerging.

Another major finding is that governments remain accountable for AI-driven public services even though much of the technology is controlled by global software vendors. AI capabilities are frequently added automatically to existing Software-as-a-Service (SaaS) platforms, giving governments little control over software updates, model behaviour, or system changes. This creates a growing gap between responsibility and operational control, increasing dependence on private technology companies.

The report also highlights that many countries lack dedicated institutions capable of evaluating AI performance, monitoring risks, or investigating failures. Responsibilities remain fragmented across ministries, making governance reactive instead of preventive. As AI becomes embedded in identity systems, welfare programmes, healthcare, education, financial services, and public administration, this lack of oversight could expose governments to growing operational and legal risks.

Procurement Is Becoming the New AI Governance Tool

Rather than waiting years for comprehensive AI legislation, the report argues that governments already possess one of the most powerful governance tools: public procurement.

Instead of treating procurement as a simple purchasing exercise, governments should require AI vendors to meet clear standards before deployment. Contracts should include requirements for transparency, performance testing, audit rights, cybersecurity protections, data governance, human oversight, incident reporting, and clear accountability for system failures.

This recommendation is particularly important because procurement decisions are often made long before AI systems begin affecting public services. Embedding governance requirements at the purchasing stage allows governments to reduce future risks while encouraging vendors to build safer and more transparent products.

The report also recommends that procurement should evolve into a continuous governance process rather than a one-time transaction. Governments should regularly monitor AI performance, require periodic reviews, and establish clear exit strategies that prevent excessive dependence on individual technology providers. For policymakers, this approach offers an immediate and practical pathway to improving AI governance without waiting for complex regulatory reforms.

Strong Digital Infrastructure Will Decide AI Success

The report makes it clear that AI cannot perform effectively without a strong digital public infrastructure. Countries with reliable digital identity systems, interoperable government databases, secure payment platforms, and high-quality public data are better positioned to deploy AI safely and improve services over time.

In contrast, introducing AI into fragmented digital systems can magnify existing weaknesses. Authentication failures, identity verification errors, language misunderstandings, fraud detection mistakes, or system outages can quickly prevent citizens from accessing welfare payments, healthcare, education, banking, or legal services. In many developing countries, correcting these errors requires lengthy manual verification processes, increasing administrative costs while reducing public trust.

Language also remains one of AI's biggest challenges. Many AI systems continue to perform poorly in local languages, dialects, and multilingual environments, limiting their usefulness across much of the developing world. The report argues that governments and development partners should invest in local-language datasets, multilingual evaluation systems, and culturally appropriate AI testing to improve accuracy, fairness, and accessibility.

Importantly, the report views small states, Small Island Developing States (SIDS), and low- and middle-income countries (LMICs) as valuable testing grounds rather than exceptions. Because these countries often have limited institutional capacity and fewer financial resources, AI failures become visible much earlier. Governance models that succeed under these conditions are likely to prove effective globally.

Shared Global Action Can Build Trustworthy AI

The report concludes that responsible AI adoption cannot be achieved by governments working alone. International development partners, including multilateral development banks, UN agencies, and bilateral donors, should shift their focus from supporting AI innovation alone to strengthening institutional capacity.

Priority investments should include AI procurement standards, government risk assessment capabilities, cybersecurity, multilingual AI resources, digital public infrastructure, workforce training, evaluation tools, and regional knowledge-sharing platforms. Since many countries cannot independently build comprehensive AI assurance systems, regional cooperation can reduce costs while improving governance standards and bargaining power with technology vendors.

For the private sector, the findings signal both opportunities and responsibilities. Technology companies that provide transparent, auditable, secure, and locally adaptable AI solutions are likely to gain greater trust from governments. Demand is also expected to grow for AI auditing firms, cybersecurity providers, cloud infrastructure companies, digital identity specialists, language technology developers, and AI assurance services.

Ultimately, the report argues that AI governance should no longer be viewed solely as a regulatory challenge but as an investment in institutional resilience and long-term economic development. Countries that strengthen procurement systems, improve digital infrastructure, build technical expertise, and establish effective accountability mechanisms today will be better positioned to attract responsible investment, improve public services, protect citizens, and ensure AI contributes to inclusive and sustainable development. As AI becomes increasingly embedded across public institutions, the quality of governance, not just the quality of algorithms, will determine whether societies fully realize the technology's economic and social benefits.