Global adoption of artificial intelligence continued to grow steadily during the second half of 2025, reaching 16.3% of the world’s population, which is equivalent to approximately one in six people now using generative AI tools to learn, work or solve problems. The advance, of 1.2 percentage points compared to the first half of the year, is significant for a technology that has barely entered mass use in recent years and confirms that AI has become an everyday infrastructure in broad layers of society.

The data, which is extracted from the “AI Diffusion Report 2025” report prepared by Microsoft, reflects, however, that progress is not distributed homogeneously: adoption in the so-called Global North grew almost twice as fast as in the Global Southwidening the existing gap. By the end of 2025, 24.7% of the working-age population of the Global North used AI tools, compared to 14.1% in the Global South. The difference, far from being reduced, went from 9.8 to 10.6 percentage points in just six months, showing that the benefits of AI are expanding, but they are not doing so equitably.

This pattern confirms that access to AI does not depend solely on technological availability, but on decisions accumulated over time in terms of digital infrastructures, public policies, education and social trust in technology, as noted in the report. The countries that lead the ranking are not necessarily those that develop the most advanced models, but rather those that have systematically integrated AI into their economy and daily services.

The report places the United Arab Emirates at the top of the world ranking, because it reached an adoption rate of 64% of its working-age population in the second half of 2025.
The case of the United Arab Emirates is especially illustrative. His leadership is not temporary, but the result of a strategy initiated in 2017, when the country appointed the world’s first Minister of State for Artificial Intelligence and launched a national AI strategy with clear sectoral priorities. When the wave of generative AI reached the general public, the technology was already part of citizens’ daily lives, which also explains the high levels of trust: the Edelman Trust Barometer places trust in AI in the Emirates at around 67%, compared to the 32% registered in the United States or the Western European average.

In contrast, the United States offers a relevant paradox. Despite leading the development of frontier models and global technological infrastructure, the country fell from 23rd to 24th place in relative adoption during the second half of 2025, with a usage rate of 28.3%. The data suggests that innovation alone does not guarantee widespread adoption if it is not accompanied by integration, training and standardization policies for use.

South Korea: the clearest acceleration of 2025

Among the countries analyzed, South Korea stands out as the most dynamic case of the year. In just six months, it climbed seven positions in the global ranking, from 25th to 18th, registering the largest increase in adoption of all the countries analyzed. The use of generative AI exceeded 30% of the population, with cumulative growth of more than 80% since the end of 2024, well above the global average and also the growth recorded in the United States.

This jump responds to the convergence of three factors. First, an intensification of public policy, with the creation of the National AI Strategy Committee, the approval of the “AI Basic Act” and new investments in education and talent in regions outside large urban centers. Second, substantial improvements in the performance of Korean language models. The leap in quality between GPT-3.5, GPT-4o and GPT-5 transformed AI into a reliable tool for complex tasks in the local language, from education to professional use.

The third catalyst was cultural. In April 2025, the viral Ghibli-style image generation with ChatGPT-4o spread massively on Korean social networks, introducing AI to millions of new users without technical barriers. Usage data suggests that a relevant portion of those users continued to explore other features, turning a viral phenomenon into sustained adoption. South Korea thus consolidated itself as one of the fastest growing ChatGPT markets, to the point of motivating the opening of an OpenAI office in Seoul.

DeepSeek and the role of access in the geopolitics of AI

One of the most unexpected developments of 2025 was the rise of DeepSeek, an open source AI platform that achieved rapid adoption in markets historically underserved by large Western vendors. Their strategy combined three elements: release of the model under the MIT license, a completely free chatbot, and active distribution in regions with economic or political barriers to accessing American services.

DeepSeek adoption was especially high in China, Russia, Iran, Cuba and Belarus, but also in Africa, where its use is estimated to be two to four times higher than in other regions. In the latter case, the expansion was reinforced by alliances with Chinese technology companies and telecommunications operators, which highlights how access, price and distribution can be as determining as the technical quality of the model.

The rise of DeepSeek introduces a clear geopolitical dimension to the global spread of AI. The competition between the United States and China is not limited to the development of models, but extends to the ability to promote their adoption in different regions of the world. At the same time, it raises questions about governance, standards and security, since open models can scale quickly with more limited control mechanisms.

The global balance drawn by the report is that of a world that adopts artificial intelligence at a remarkable speed, but in which structural inequalities become more visible. Countries with solid infrastructure, coherent policies and linguistically adapted models consolidate their advantage, while large regions of the Global South They advance at a slower pace, conditioned by economic, educational and access barriers.

As Microsoft points out, the next phase of the AI ​​revolution will not depend only on technical advances, but on strategic decisions about access, education, trust and governance. The challenge is not only to innovate faster, but to ensure that this innovation benefits more people and more economies, preventing the digital divide from becoming a structural gap that is difficult to close.

More info: AI Diffusion Report 2025