Who gets to govern AI? U.N. panel warns access alone is not enough
Preliminary report argues that countries may gain access to artificial intelligence while losing influence over the standards, infrastructure and governance systems that shape it.
A new United Nations scientific assessment warns that artificial intelligence risks creating a new form of global inequality in which countries gain access to powerful technologies while losing influence over how those technologies are developed, governed and deployed.
The warning comes in the first preliminary report issued on Wednesday by the Independent International Scientific Panel on Artificial Intelligence, a group of 40 experts established by the U.N. General Assembly to provide scientific assessments of AI’s opportunities, risks and impacts.
The report depicts a world in which artificial intelligence is advancing rapidly while the capacity to govern it remains concentrated in a handful of countries and companies. The result, the panel warns, could be a widening gap not only in access to AI technologies but also in the ability to shape their future development.
“This report is the moment the world stops being able to say we did not know,” Nobel Peace Prize-winning journalist Maria Ressa, who co-chairs the panel with computer scientist Yoshua Bengio, told a U.N. press briefing. “No expert today can promise that the most advanced systems will do what you instruct them to do,” she said. “This report is the common ground to act from.”
While much public debate has focused on AI safety, jobs and economic disruption, the report highlights a different concern: the growing concentration of AI capabilities in a small number of countries and companies.
“Access to AI tools alone does not produce equal benefit,” the report says. “Countries that rely on foreign models, cloud infrastructure and data pipelines may gain access to AI while losing practical control over its standards, safeguards and local fit.”
The finding points to one of the central governance questions of the AI era: who gets to shape the rules, standards and institutions surrounding technologies increasingly embedded in economic, political and social systems.
The report notes that the United States and China dominate the development of the most advanced AI models and much of the computing infrastructure required to train and deploy them. It warns that the concentration of AI capabilities in a small number of firms and countries could undermine democratic accountability and leave many nations dependent on technologies they neither control nor fully understand.
Bengio said the report is meant to ensure that policy decisions are informed by the best evidence available about how to establish global governance for “the growing intelligence of machines — and you have to realize that intelligence gives power.”
The concentration extends beyond companies to infrastructure. The report estimates that the United States accounts for roughly 75% of computing power among the world's 500 largest AI computing clusters, compared with 15% for China and 10% for the rest of the world combined.
“The gap between the haves and the have-nots is likely to grow, and I don’t think we’re taking the full measure of that,” Bengio said.
The report also notes that institutions based in the United States produced 59 notable AI models in 2025, compared with 35 in China and 13 across the rest of the world.
The imbalance is particularly significant because many governments lack the expertise needed to evaluate rapidly advancing AI systems.
“Most countries, including many advanced economies, lack the technical expertise to assess the most capable ‘frontier’ models or to participate meaningfully in their governance,” the report concludes.
More than 1 billion people now use conversational AI each week, according to the assessment, while the most advanced systems have achieved rapid gains on scientific and mathematical benchmarks.
One test designed to challenge frontier AI models saw top scores rise from 8% to 45% in just 16 months, while leading systems now correctly answer roughly 95% of questions on a PhD-level scientific reasoning benchmark.
The report surveys AI developments across seven areas, including scientific advances, economic impacts, security risks, human rights, democracy and governance. It argues that AI already offers significant opportunities in fields such as healthcare, education, agriculture and scientific research, while simultaneously creating risks that extend from fraud and cyberattacks to disinformation and political manipulation.
The panel warns that AI capabilities are improving faster than the evaluation methods, oversight mechanisms and governance frameworks needed to assess and manage the technology's risks. One chart included in the assessment shows frontier AI capabilities improving nearly twice as fast since April 2024. Another indicates that the complexity of tasks AI systems can complete autonomously has been doubling every 4.6 months since late 2024, a significant acceleration from earlier trends.
The assessment arrives as governments prepare to gather in Geneva next week for the inaugural U.N. Global Dialogue on AI Governance, where the panel’s findings will be presented to member nations. The meeting is part of a broader effort by the United Nations to build international mechanisms capable of responding to the accelerating development of artificial intelligence.
Speaking at the report’s launch with Ressa and Bengio, U.N. Secretary-General António Guterres described the panel as “the first global, fully independent scientific body” dedicated to closing the world’s AI knowledge gap and assessing the technology’s impacts across economies and societies. He urged governments not to delay action as AI capabilities continue to advance.
“The more AI advances without shared rules, the less say governments and people will have in the outcome,” Guterres said. "The science is here. We can no longer say we did not know."
Facts, truth and trust
The report also highlights disparities in language access and digital infrastructure.
While generative AI systems perform well in English and other widely used languages, many languages remain poorly represented in training data and digital tools. More than 2 billion people worldwide remain entirely offline, limiting their ability to participate in AI-driven economic and social transformations.
The governance gap is similarly pronounced. According to the report, 118 countries, most of them in the Global South, are not participating in major AI governance discussions, while fewer than one-third of developing countries have adopted national AI strategies. The panel argues that countries lacking computing infrastructure, technical expertise and testing capacity risk becoming consumers of AI systems rather than participants in shaping them.
Guterres said the panel's findings underscore the need for broader participation in AI governance, arguing that countries must have the capacity not only to use artificial intelligence but also to help shape the rules governing it. He said he plans to present proposals aimed at helping nations build that capacity and share in the technology's benefits.
The panel argues that these disparities risk widening existing divides unless governments invest in computing infrastructure, AI literacy, technical expertise and local AI ecosystems.
Yet some of the report’s strongest language appears not in the assessment itself but in a separate message from Bengio and Ressa. “We are deeply concerned,” they wrote. “The race for competitive advantage is leading the world into risks whose magnitude is grossly underestimated.”
The two argued that AI development is increasingly concentrated among a handful of governments and companies, raising questions not only about technological leadership but also about political power.
“Concentration of capability is becoming concentration of political power,” they wrote. “The sovereignty of most nations is at stake.”
The report does not prescribe specific policies. Its mandate is scientific rather than political, providing governments with a shared evidence base rather than negotiating outcomes. But both the report and Guterres' remarks point toward the same conclusion: the next phase of AI governance may depend less on access to technology than on the ability of countries to understand, evaluate and influence the systems increasingly shaping their futures.
Ressa, whose journalism career has focused on disinformation and democratic resilience, framed AI governance as a challenge to information integrity itself. “Without facts, you cannot have truth. Without truth, you cannot have trust,” the co-chairs wrote. “Without these three, we have no shared reality.”
