China Has Erased the AI performance Gap With the U.S., Said the Stanford HAI Report

Public trust in AI oversight hit a new low while this technology is being adopted at a record-breaking pace, said researchers from Stanford University’s 2026 AI Index Report. The recently released report covers the biggest technical advances, investments, and trends in education, health, legislation, and the environment, offering an empirical foundation about AI’s rapid evolution and real-world adoption. Known as Stanford HAI, this report, now into its ninth year, is a comprehensive annual study by the Stanford Institute for Human-Centered Artificial Intelligence. The report found that the adoption of generative AI has grown faster than any other technology in history, with 53% of the world’s population now using it regularly. Opinions on the technology are mixed: 59% say it provides more benefits than drawbacks, while 52% say it makes them nervous. This year, one of the most striking takeaways is the race for global dominance, with China having erased the AI performance gap with the U.S., leaving them neck and neck. The U.S. maintains a significant edge in terms of capital, infrastructure buildout, and AI chips, but China now holds sway in other key areas, such as patents, publications, and autonomous robotics development, or “physical AI.” Other nations, such as South Korea, have emerged as the world’s leaders in “innovation density,” filing more patents per capita than any other country. For many governments in Europe and Central Asia, AI infrastructure sovereignty has become a top policy priority. South American and Middle Eastern nations lag far behind, and this could lead to a new kind of “digital divide,” More than 90% of all notable AI models are now created by private companies — such as Google LLC, Anthropic PBC, and OpenAI —, spreading concerns about AI “black boxes”. The presence of neutral academics has plummeted. These AI leaders have all abandoned the practice of disclosing the dataset sizes and training durations of their latest models. Moreover, 80 of the 95 most notable models launched last year were released without their training code. Only 31% of U.S. citizens now trust their government will regulate AI properly. In China, 27% of people trust their government, and in the EU, 53% express confidence.