Chloé Maurel
Since the early 2020s, artificial intelligence (AI) has become a major issue in global governance, raising economic, environmental, and ethical challenges. The rapid advances of machine learning systems - particularly generative models and autonomous systems - has given rise to new political, economic, and societal risks: manipulation of information, automation of warfare, mass surveillance, the amplification of technological inequalities, and systemic risks linked to advanced systems.
An energy-intensive technology
The rise of artificial intelligence has led to a sharp increase in energy consumption. In 2024, data centers consumed approximately 415 TWh of electricity, or nearly 1.5% of global consumption - equivalent to the annual electricity use of a country like France. This demand is growing very rapidly. Training large-scale models is particularly energy-intensive: a single very large model may require tens of gigawatt-hours, equivalent to the annual consumption of tens of thousands of households. This comes at a time when millions of people worldwide still lack access to electricity for their daily lives. Data centers also consume large amounts of water for cooling: AI could require more than 6 billion cubic meters of water by 2027. Overall, digital infrastructures related to AI already account for around 0.5% of global CO₂ emissions, a figure that is steadily increasing.
That said, there is not necessarily any reason to be categorically opposed to AI as a new technology. Just as the invention of the steam train and the automobile did not eliminate walking, just as the invention of cinema did not eliminate theatre, just as the internet and digitalisation did not eliminate libraries and printed books, and just as the calculator did not eliminate mathematical reasoning — so too may AI, if used intelligently and judiciously, not necessarily displace human reasoning and intelligence, but could even stimulate them.
Yet the adverse effects of AI are plain to see: nearly 40% of jobs worldwide are already threatened by AI, particularly in administrative, legal, translation, and creative fields.1 This proportion could soon reach 60% of all employment.
For AI to play a positive role in serving the entire global population, and for its negative impacts on humanity to be minimized, it is essential to establish global regulation of this new technology. Yet the body best suited to achieve this is the most universal international organization, bringing together 193 member states on an equal footing—namely, the United Nations. How can this be achieved?
A Major Issue for the United Nations
The stated ambition of UN Secretary-General António Guterres is to build a multilateral framework capable of governing AI while enabling its development in the service of human progress. This ambition is embedded in a progressive vision of global governance — one of an international system capable of producing global norms to bring this new technology under control.
An Unequal Concentration of Technological Power
The emergence of AI as an international issue stems from a profound transformation of the technological and geopolitical landscape. Artificial intelligence systems are currently developed mainly by a few major technological powers and private companies, particularly in the United States and China. This concentration of technological power raises the issue of a global imbalance in access to digital resources, which can be described as an “AI divide.” In addition, the military applications of AI—such as autonomous weapons systems or automated cyberattack tools—are fueling growing concerns about a new technological arms race.
A Problematic Dilution of Responsibility in Warfare
As legal analyst Ridwane Allouche argues, “one of the principal dangers of military AI lies in the gradual diffusion of responsibility.” Indeed, “who” is responsible when a drone or robot activated by AI inflicts death? “The soldier who validates the recommendation? The engineer who designed the model? The company that trained it? The state that deployed it?"2 The laws of war, as defined notably by the Geneva Conventions (1949), will therefore need to be rewritten and updated to account for the new reality of military AI.
In this context, the absence of a specific international legal framework for AI constitutes a major problem. Existing regulations are mainly national or regional, such as the European regulation on artificial intelligence (“AI Act,” adopted in 2024), while other powers favor more flexible approaches. This normative fragmentation risks producing an incoherent international system, where rules differ depending on regions of the world. The UN therefore appears as the most legitimate institutional space to attempt to build a universal regulatory framework.
The UN Secretary-General has thus emphasized the need for global AI governance comparable to that established for other high-risk technologies, such as nuclear energy or biological weapons. The aim is to legally fill a “dangerous void,” considering that “the power of AI is too great to be left in the hands of a privileged few.”
