Summary

The UN Independent International Scientific Panel presented a report on the risks and opportunities associated with artificial intelligence. AI can contribute to achieving the Sustainable Development Goals; however, the pace of its development and deployment is outstripping existing capacities for assessment, regulation and oversight. More than one billion people worldwide use AI every week, yet access to infrastructure, data and expertise remains highly unequal. The development of frontier models is concentrated among a limited number of countries and companies. In 2025, 59 frontier AI models were developed in the United States and 35 in China, compared with only 13 in all other countries combined. At the same time, 91% of such models were developed by the private sector. Key risks include the spread of disinformation, an increase in cyberattacks, harm to mental health and widening digital inequality. Particular attention is given to agentic AI — systems capable of independently planning and carrying out tasks with minimal human involvement. Such technologies have already accelerated the discovery of new materials by more than tenfold and reduced the workload associated with scientific literature reviews by approximately 60%. At the same time, reliable mechanisms for AI oversight have yet to be established. Priority tasks include independent assessment and auditing of AI systems, as well as the development of transparent accountability mechanisms.