LocationsSeeing Around the Bend
Expanding agency in the age of AI adoption
2026 is becoming a turning point for how the world responds to AI.
Following the momentum generated by the India AI Impact Summit, the Hamburg Sustainability Conference, the first UN Global Dialogue on AI Governance, the AI for Good Global Summit, the WSIS Forum, Digital@UNGA 2026 and the road to the next AI Impact Summit in Switzerland all come at a moment when the terms of AI adoption are being set.
What matters now is whether global cooperation can translate into practical support for countries who need tools, standards and infrastructure that benefit everybody, not just the privileged.
AI enters the mainstream
In just three years, generative AI has reached more than half the global adoption level achieved by the internet or personal computers over much longer periods. By the first quarter of 2026, generative AI use had reached 17.8 percent of the world’s working-age population, up from 16.3 percent in the second half of 2025. Twenty-six economies now report generative AI usage above 30 percent among working-age adults.
Speed of AI adoption by technology
Source: The Project on Workforce at Harvard, 2025 | Chart: 2026 AI Index report
Usage has reached 27.5 percent in the Global North and 15.4 percent in the Global South, and the gap is widening. An estimated 50 times more ChatGPT users are in high- rather than in low-income countries. Fewer than one in five workers in low-income countries are exposed to AI.
Source: World Bank Data360
Source: World Bank Data360
Yet these figures still don’t tell us enough about who shapes the terms on which AI enters economies and institutions.
The trajectory of AI is not predetermined. The choices being made today on data, infrastructure, governance, skills, procurement and public oversight will shape whether AI expands opportunity and human agency or reinforces existing inequalities.
This is where the arc bends: as AI moves from frontier debate into institutions, services, markets, infrastructure, data systems and everyday decisions.
The question is no longer simply how fast AI spreads, but who influences how it is used and who benefits from its adoption.
Expanding choices
Bending the AI Arc Towards Equity and UNDP’s 2025 Human Development Report, A Matter of Choice: People and Possibilities in the Age of AI, have placed humans at the centre of what could be one of the most profound transitions of our lifetimes.
AI can foster development when countries have the foundations to adapt it to their needs. Data shapes whether systems reflect the people, places, languages and realities they are meant to serve.
Insights from UNDP’s country engagements reveal a practical picture. In many places, AI is entering institutions through software platforms, procurement decisions, vendor tools and everyday use before the systems needed to coordinate, govern, evaluate and sustain it are fully in place.
But when those foundations are strong, AI can help governments improve public services, create decent work and open new opportunities.
Agency is not full technological self-sufficiency. For many countries, the practical goal is to build enough domestic capability to make choices within an interdependent global AI ecosystem.
UNDP’s focus
UNDP’s Strategic Plan 2026-2029 positions digital and AI transformation as key accelerators for development.
Across its work in more than 170 countries and territories, UNDP increasingly sees AI evolving from a technology conversation into an implementation and systems challenge. The outcomes AI produces are shaped not only by the technology itself, but by the institutions, governance arrangements, data ecosystems, talent, financing and digital public infrastructure that surround it.
UNDP is helping countries understand AI, what choices it creates and what foundations are needed to make it safe, inclusive and useful for everybody.
The AI Sprint initiative supports more than 60 interventions in over 40 countries, including AI landscape assessments, national AI strategies, trust and safety support, local-language ecosystems, governance support and public-sector capacity-building.
Through AI Landscape Assessments, UNDP supports countries in understanding how AI can support national development priorities and what foundations are needed to get there. In Malawi, the assessment connected AI opportunities to the practical foundations needed for responsible adoption, including data stewardship, skills, compute and safe testing. In Guatemala, it supported government efforts to assess both the promise of AI and the harder questions of how to govern it.
For some countries, the work feeds straight into national strategies. In Mauritius, UNDP helped the government run a multi-stakeholder consultation that mapped out a practical path for adopting AI across priority sectors. In Zimbabwe and Ghana, UNDP’s support in shaping national AI strategy was paired with hands-on training to help government officials put policy into practice.
The launch of Mauritius’ National AI Strategy and FAIR Guidelines marks a key milestone in building an ethical, inclusive and people-centred AI ecosystem, with technical support from UNDP.
Elsewhere, UNDP’s support has addressed distinct challenges in different contexts. In Ecuador, it charted a roadmap toward an AI regulatory sandbox. In Rwanda, AI trust and safety support surfaced the need for clearer institutional coordination around AI risk and escalation. In Trinidad and Tobago, procurement emerged as a practical lever for steering AI adoption in a fast-moving environment.
While the conventional AI narrative centres on tools, models and infrastructure, an underrecognized bottleneck to AI's positive impact in government is human capacity. A country can have high-end data infrastructure and a national AI strategy but still fail to act on either, because the officials who hold decision power do not have the conceptual language, the critical lens, or the hands-on experience to govern, commission, or deploy AI responsibly. That is the gap UNDP's work fills. Most AI capacity building for government is shallow awareness-raising. What UNDP is building is more ambitious: officials who can interrogate AI systems, co-design responsible applications and lead institutional transformation from the inside.
UNDP is also exploring how AI can support concrete development challenges, from strengthening policy coherence in Indonesia and countering digital scams to improving anticipatory action, disaster response and resilience planning in crisis and fragile settings.
Additionally, UNDP is helping build multistakeholder partnerships around responsible AI at the global level. Through the Hamburg Declaration on Responsible AI for the SDGs, launched in 2025, UNDP is supporting a growing coalition committed to translating responsible AI principles into practical action. The initiative demonstrates how multistakeholder partnerships can help advance inclusive AI adoption, strengthen accountability and mobilize collective action around shared development goals.
Through the G7-endorsed AI Hub for Sustainable Development co-led by UNDP and the Ministry of Enterprises and Made in Italy, we are re-imagining global AI partnerships and powering local ecosystems with Africa. Focused on unlocking resources for today and building foundations for tomorrow, this work is anchored in 18 partner countries with G7 partners reaffirming their commitment with the May 2026 adoption of the G7 Digital and Technology Ministerial Declaration under the French G7 Presidency.
UNDP also plays a leading role in advancing international cooperation on AI across the United Nations. Through its leadership in interagency collaboration, policy dialogue and capacity-building initiatives, UNDP helps connect global AI governance discussions with the practical needs of countries, translating emerging frameworks and principles into solutions, guidance and support that governments can apply in practice.
This support includes advancing shared approaches to AI readiness, governance, institutional capacity-building and knowledge exchange, while contributing to emerging global mechanisms such as the UN AI Resource Hub that aim to make expertise, tools and learning more accessible across countries and partners.
Examining how AI affects people
Institutions cannot govern what they cannot see.
As AI becomes embedded in public services, finance, information ecosystems and administrative processes, the risks move closer to people’s lives. AI-enabled scams, algorithmic bias, data exploitation and misinformation weaken public trust. A flawed system can delay a social protection payment, misread a person’s identity, exclude someone from a service, distort public information, or make it harder for people to understand and challenge decisions that affect them.
This is especially visible in Small Island Developing States and other resource-constrained countries. They often show earlier and more clearly what many countries will face as AI adoption grows.
Most countries will not build frontier models. But they will still be responsible for how AI systems affect their people. Many will rely on systems designed elsewhere, governed elsewhere, and optimized for realities other than those of the people and institutions expected to rely on them.
That is why trust and safety need to be built into AI adoption from the start. Countries need practical ways to ask better questions before systems are adopted, monitor how tools perform, respond when something goes wrong, and make sure people can challenge or correct outcomes.
Governance lives in these everyday choices: what governments buy, how public services use data, how technology providers are held accountable, and whether people know where to turn when an AI-enabled system affects them unfairly.
What gives countries room to act
Access to AI is not the same as agency in how it is adopted, adapted and used.
Language is one of the clearest examples. Of the more than 7,000 languages, AI effectively supports fewer than 100. Most African languages, along with many in South and Southeast Asia, the Pacific and Indigenous communities, remain underrepresented.
Which means people may have AI tools without receiving the benefit. Systems may misunderstand needs, overlook local knowledge or leave communities out of next generation digital services and economic opportunity.
Language inclusion is about whether communities are visible. It is also about whether value flows back to the people whose knowledge, language and data make better AI possible.
The same is true across other foundations. Data, compute, talent, governance, digital public infrastructure, trust and safety and institutional capability are not just technical building blocks. They shape the choices available to governments, innovators, public institutions and communities.
A country can have AI and still have limited room to influence how it is used.
Agency is created when countries can connect public needs, local capability, trusted partnerships, safeguards and practical implementation capacity.
The future of AI may depend as much on bringing intelligence closer to users and communities as on building larger centralized systems.
Seeing around the bend
The future of AI is still open. But it is already being shaped where people learn, work, access services, build livelihoods, participate in public life and respond to crisis.
AI will not advance human development on its own. That will depend on whether countries have the foundations to use AI well, the institutions to govern it responsibly, the systems to create local value and adapt to local needs, and the safeguards to protect people when systems fail.
Seeing around the bend means paying attention to what adoption is already showing us and helping countries act before today’s choices become harder to change.
The signals are already visible. The question is not only who has access, who builds, or how fast AI moves. It is whether people and countries have the agency to shape what comes next.
