
Image created by artificial intelligence.
Artificial intelligence is now making its presence felt in every sphere of society, from schools to hospitals and from social media to weapons systems. In the face of this acceleration, UN Secretary-General António Guterres believes that the world can no longer make do with fragmented regulation. Opening the first Global Dialogue on AI Governance in Geneva on Monday, he called on states to build an international architecture capable of governing a technology whose effects now transcend national borders.
“Machines can inform decisions, but human beings must decide… and be held accountable,” the UN chief said, making the case for global governance “worthy of everyone’s trust”.
The message found an immediate echo with the UN Secretary-General’s Envoy on Technology, Amandeep Singh Gill. “Artificial intelligence is too important to be shaped by a handful of players,” he said, calling for a global dialogue that is “inclusive and evidence-based”.
A global pact to protect children
The most striking announcement of this first day, however, concerned the youngest. António Guterres proposed the adoption of a global commitment on children’s safety in the face of artificial intelligence, which could become one of the first pillars of a future international framework on AI.
The text championed by the Secretary-General notably provides for a ban on making artificial intelligence tools available to minors until they have undergone an independent safety assessment. It also calls for a zero-tolerance policy towards AI-generated child sexual abuse material and demands that chatbots systematically direct children towards human help whenever they detect signs of distress or psychological suffering.
“No child should be a guinea pig for unregulated artificial intelligence,” António Guterres insisted. To illustrate his point, he drew a parallel with the safety standards applied to medicines and toys. “We do not allow a medicine to be given to a child until its safety has been proven. We test every toy. Yet AI has already entered our children’s lives – their learning, their friendships, their most private questions – before anyone has even asked what effects it might have on them.”
The President of the United Nations General Assembly, Annalena Baerbock, also warned of risks that are already visible. Citing recent studies, she noted that 99% of deepfakes circulating online are sexual in nature and that 96% target women or girls, illustrating the new forms of violence made possible by generative tools.
A technology advancing faster than the rules
Beyond child protection, UN leaders stressed the widening gap between the speed of technological progress and the ability of states to regulate it. In less than three years, generative AI has upended economies, public administrations and the daily lives of hundreds of millions of people. But this rapid spread brings mounting risks: large-scale disinformation, cyberattacks, automated surveillance, algorithmic discrimination and even loss of control over the most advanced systems.
For Yoshua Bengio, co-chair of the independent scientific panel mandated by the UN General Assembly, the capabilities of the models will continue to advance at a rapid pace. “Extremely worrying tests have shown that the most advanced AI models are capable of deceiving human beings and of understanding that they are being evaluated,” he explained.
In his view, these developments could profoundly alter the international balance of power. “It sounds like science fiction, but it is a very real possibility. It could transform the world in ways we do not yet understand.”
The findings of the scientific panel’s first report, published the previous week, point in the same direction: the experts consider that the development of AI is now outpacing scientific understanding of its risks, as well as the ability of governments to adapt their regulatory frameworks.
Avoiding a global artificial intelligence divide
The Secretary-General also stressed the need to prevent a new technological divide from emerging between developed and developing countries. “We cannot allow the digital divide to turn into an artificial intelligence divide, and that divide to then become a chasm in development, security and sovereignty,” he warned.
More than 20 states now support a United Nations initiative to create a global cooperation network to help the least developed countries build their capacities in the field of AI.
The environmental challenge
The UN also wants companies in the sector to report on the environmental impact of their systems. António Guterres called for major companies to systematically disclose the carbon emissions, water consumption and land use associated with their AI models, while also calling for data centres to run exclusively on renewable energy by 2030.
According to him, by that date artificial intelligence infrastructure could consume more electricity than every country in the world bar five, and use enough water to meet the needs of the entire population of sub-Saharan Africa for a whole year.
Towards an international law of AI
The Global Dialogue held in Geneva on 6 and 7 July marks a new stage in the United Nations’ efforts to build global governance of artificial intelligence. Following the creation of an independent scientific panel in 2023 and the adoption, in 2024, of the Pact for the Future and the Global Digital Compact, the meeting opens a new cycle of negotiations bringing together governments, technology companies, researchers and civil society organisations. A second dialogue is already scheduled for May 2027 in New York.
For the United Nations, the objective is now clear: to bring artificial intelligence gradually within the scope of international law, before technological developments make any regulation impossible.
Translated from French. The French original prevails.






