The rapid adoption of artificial intelligence brings not only technological progress but also serious risks to the global economy and social justice. The latest preliminary report from the UN Independent International Scientific Panel on AI, prepared by 40 independent experts, clearly indicates that access to AI tools alone does not guarantee equal benefits for all countries.
The key problem I see as an analyst is the unprecedented concentration of computing infrastructure and advanced model development. According to the report's data, the United States controls about 75% of all computing power of the world's leading AI supercomputers, while China controls approximately 15%. Together, these two countries accumulate about 90% of all computing used to train the most powerful AI systems. This creates a situation where countries dependent on foreign models and cloud infrastructure gain access to the technology but lose practical control over its standards, safety mechanisms, and adaptation to local conditions.
Risks to democracy and language inequality
The report rightly notes that the concentration of AI capabilities among a small number of companies and states could strengthen authoritarian control and weaken democratic accountability. Lagging countries are strongly advised to develop their own computing and data infrastructure, train specialists, and support local developers.
The issue of language inequality deserves special attention. Generative models demonstrate significantly better performance with English and other widely spoken languages, while less common languages are either barely represented or receive noticeably poorer quality. The example cited in the report of machine translation from Tigrinya — where the system confused smallpox with syphilis, gonorrhea with diabetes, and translated a phrase about intravenous antibiotics as "intravenous insecticides" — vividly demonstrates that such errors can be literally deadly.
Fundamental problems and development strategies
A separate and perhaps the most fundamental factor of inequality is basic internet access. According to the International Telecommunication Union, in 2025, 2.2 billion people remain offline. For these regions, AI adoption first hinges on access to the internet, energy, and digital skills.
The report is not a regulatory document but offers governments a scientific assessment of risks and opportunities. Among the key areas highlighted by the authors are: developing local AI infrastructure and data centers, improving AI literacy in schools and workplaces, investing in developers and technical expertise, creating AI safety institutions, strategies to combat disinformation, and mandatory testing of AI systems after deployment on real users, tasks, and conditions.
As a professional analyst, I believe the UN report timely raises a key issue: the uneven distribution of AI capabilities could lead to a technology meant to unite the world becoming a new tool for dividing it. The gap between rapidly growing AI capabilities and the lack of effective risk management methods could indeed lead to catastrophic consequences if coordinated international action is not taken.