UN panel warns AI could widen global inequality

The United Nations’ independent scientific panel on artificial intelligence is warning that the rapid spread of AI could deepen global inequality unless countries build local capacity and agree on common rules.

The panel’s report, published on July 1, says access to AI tools alone does not guarantee equal benefit. Countries that depend on foreign models, cloud infrastructure and data pipelines may end up using systems they do not control, while the gains from AI continue to concentrate in a small number of places.

UN Secretary-General António Guterres said governments should not wait for the science and warned that more AI progress without shared rules would leave governments and people with less say over the outcomes.

A warning ahead of Geneva

The report was released ahead of the UN’s inaugural global dialogue on AI governance in Geneva next week. That timing gives the panel’s warning immediate political weight, because member states are already moving toward discussions on how AI should be governed internationally.

The report frames AI as a technology with real upside, including for education and agriculture, but also as one that can intensify fraud, election interference, authoritarian control and other harms if deployment and oversight remain uneven.

The panel argues that the question is not whether AI will spread. It is who will shape it, who will be able to evaluate it and who will end up dependent on systems built elsewhere.

Who is most exposed

The panel says the burden will fall most heavily on countries that rely on outside providers for core AI infrastructure. That includes places dependent on foreign models, cloud services and data pipelines, as well as regions with weaker internet access and limited language coverage.

The report also says adoption in the global south is lagging behind the global north. That gap is not only about access to software. It is also about the ability to build and maintain the institutions needed to use AI safely and effectively.

According to the panel, many countries, including some advanced economies, still lack the technical expertise needed to assess frontier models or govern them meaningfully.

The concentration problem

The report says the United States and China dominate leading model development and AI compute investment, giving them outsized influence over the direction of the field.

That concentration matters because the countries and companies with the most compute, data and talent are also the ones most able to set defaults, define standards and move fastest on deployment. For countries without comparable capacity, the risk is that AI becomes something imported rather than something locally shaped.

Maria Ressa, identified in coverage as a co-chair of the panel, said AI’s pace is not slowing, the power is concentrating and control is not guaranteed.

The report also raises environmental concerns tied to datacenters, including energy and water use, adding another layer to the debate over who benefits from AI and who bears its costs.

What the panel wants governments to do

The panel is urging countries to invest in local AI infrastructure, AI literacy and developer training so that adoption does not remain limited to a few global hubs.

It also calls for AI safety institutes, stronger defenses against disinformation and ongoing monitoring after systems are deployed. The emphasis is on building practical capacity, not just issuing broad statements about principles.

Taken together, the recommendations amount to a sovereignty argument: countries cannot depend entirely on foreign AI stacks and expect to retain full control over how those systems behave in their own societies.

The panel’s view is that local infrastructure and local expertise are necessary if governments want to set standards that reflect their own needs, risk tolerance and public interests.

Why the debate matters now

The report lands at a moment when AI policy is moving from broad concern to institutional design. Governments are under pressure to decide whether they want shared rules that distribute control more widely, or a future in which a few dominant firms and states define the practical boundaries of the technology.

For countries outside the current AI leaders, the issue is not just whether they can use the tools. It is whether they can evaluate them, adapt them and keep enough oversight to avoid becoming dependent on systems they do not fully understand.

For global governance, the UN panel’s warning is a push to act before the market hardens around a small set of actors. The Geneva dialogue next week is expected to be one of the first tests of whether that warning leads to any concrete commitments.

The broader stakes are straightforward: AI could become a broad productivity tool, or it could deepen the digital and economic divide that already separates countries with advanced infrastructure from those without it.

The report says the outcome will depend on choices being made now, not later.