AI is advancing faster than expected

The fastest-growing technology in history

Generative AI has become the fastest-adopted technology in history, reaching 53% of the population in just three years. By comparison, personal computers and the internet took approximately 7 to 15 years to reach similar levels of adoption. Corporate adoption has surged to 88%, while among students the technology has already become standard:

“4 in 5 university students now use generative AI.”

At the same time, the industry has largely monopolized AI development, with more than 90% of advanced models created by private companies. Transparency is declining, as companies such as OpenAI, Anthropic and Google increasingly withhold key details about their models, including data volumes and computational resources.

Technology outpaces society

The gap between technological capabilities and societal readiness is becoming a central challenge.

“It reveals a field that is scaling faster than the systems around it can adapt.”

AI systems are already demonstrating performance comparable to or exceeding human levels in several domains. Modern models can solve PhD-level scientific problems, complex mathematical tasks and programming challenges. On the SWE-bench Verified benchmark, performance rose from 60% to nearly 100% of the human baseline within a year.

However, progress remains uneven. For example, the Gemini Deep Think model achieved a gold medal-level result at the International Mathematical Olympiad, yet such systems correctly read analog clocks only 50.1% of the time. Progress is also evident in autonomous AI agents, whose success rate in completing real-world computer tasks increased from 12% to around 66%, though they still fail in roughly one out of three cases.

Rising risks

As capabilities grow, so do risks. The number of recorded AI-related incidents rose to 362, up from 233 a year earlier. Challenges are particularly visible in autonomous systems, where AI agents continue to make errors in about one-third of real-world tasks despite rapid improvements.

Concerns are also growing over the imbalance in development. The report highlights that improving one aspect of AI can degrade another. For instance, enhancing safety may reduce model accuracy. Developers continue to publish performance metrics, while safety-related data remains limited:

“Reporting on responsible AI benchmarks remains spotty.”

Economy and labor market

The economic impact of AI is already measured in hundreds of billions of dollars. In the United States alone, the value of generative AI is estimated at $172 billion annually. Productivity gains of 14–26% have been observed in areas such as software development and customer support.

However, this growth is accompanied by concerns in the labor market, as employment among software developers aged 22 to 25 has declined by nearly 20%, despite overall industry expansion.

The environmental cost of AI

Training advanced AI models requires massive resources, with emissions reaching up to 72,816 tons of CO2 equivalent.

The total power capacity of AI data centers has reached 29.6 GW, comparable to the peak demand of an entire region such as New York State.

Water usage is also rising. With a single large model potentially consuming volumes equivalent to the needs of up to 12 million people annually.

A global race

The United States and China have effectively reached parity in advanced model development, frequently exchanging leadership positions between 2025 and 2026.

China’s DeepSeek-R1 matched leading U.S. systems in early 2025, and as of March 2026, the gap between top models stands at just 2.7%.

The United States leads in high-end models and investment, while China dominates in publications, citations, patents and industrial robot deployment. Infrastructure is another key factor, with the U.S. hosting more than 5,400 data centers, though the global AI supply chain remains vulnerable due to reliance on a single chip manufacturer.

What comes next

Public perception remains divided, with 73% of experts expecting a positive impact of AI on work compared to just 23% of the general public.

The report concludes that humanity is entering a new phase where the key challenge is not only advancing AI but learning how to manage it:

“What we cannot yet measure matters just as much as what we can.”

Earlier, Qazinform News Agency reported that OpenAI called on governments and societies to prepare for the rapid transition to advanced AI, warning that future systems could outperform humans in many areas and significantly reshape the global economy, labor markets and everyday life.