The rapid deployment of artificial intelligence, despite its enormous potential, carries a serious threat of widening the gap between developed and developing countries. This is the conclusion reached by a group of 40 independent scientists and experts in a preliminary report by the UN Independent International Scientific Panel on AI. The key problem, in my opinion, lies not in access to the technology itself, but in control over it.
The report emphasizes that countries relying on imported models, cloud servers, and external data flows risk losing practical control over safety standards, protection mechanisms, and the adaptation of solutions to local needs. This creates a situation where the technology is available, but its application and development are dictated from outside.
Computational Hegemony and the Language Barrier
The main impact of inequality falls on the concentration of computing power. According to the report's data, the United States controls about 75% of the computing resources of the world's leading AI supercomputers, and China about 15%. Together, these two countries accumulate roughly 90% of all computing used to train the most powerful systems. Such concentration, the authors warn, not only strengthens authoritarian tendencies but also undermines democratic accountability, leaving "lagging" countries in the role of passive users.
Special attention should be paid to the so-called "language inequality." Generative models demonstrate high quality when working with English and a number of other common languages, while others are either ignored or receive unacceptably low quality. The report provides a telling example: a machine translation system for the Tigrinya language confused smallpox with syphilis, gonorrhea with diabetes, and translated a phrase about intravenous antibiotics as "intravenous insecticides." Such errors can be fatal.
Fundamental Challenges and Recommendations
Inequality is further exacerbated by the basic digital divide: according to the International Telecommunication Union, 2.2 billion people remain offline in 2025. For these regions, the implementation of AI runs into a lack of access to the internet, electricity, and digital skills. The report's authors propose a number of specific measures: developing local AI infrastructure, improving AI literacy, investing in developers, creating safety institutions, and strategies to combat disinformation. The need to test AI systems after launch on real users and tasks is also emphasized.
"In the long term, the gap between rapidly growing functional capabilities and the availability of effective risk management methods could lead to catastrophic results," the report notes.
The report also points to risks of fraud, election interference, human rights violations, and pressure on labor markets. Let me remind you that earlier, the Bank for International Settlements warned that the investment boom around AI itself is becoming a source of macro-financial risks.
My comment: This report is not just a warning, but a clear signal for investors and regulators. The concentration of computing power in the hands of two countries creates a monopoly environment where AI development will be dictated by their interests. For the crypto industry, which has historically advocated for decentralization, this is a challenge: can we offer alternative, censorship-resistant solutions for data storage and processing that do not depend on a single center? Ignoring this problem could lead to AI becoming not a tool for progress, but a new lever of control.