Stanford HAI: Expert Authority
By Javier Surasky
-HAI%20Institute.png)
This actor in the global AI ecosystem does not manufacture chips, operate the cloud, or train, on its own, the models that dominate the market. Its power is exercised through influence over other decision-making actors, as one of the most important reference points for understanding artificial intelligence as a global phenomenon.
Stanford HAI, the Institute for Human-Centered Artificial Intelligence at Stanford University, was created in 2019 to advance research, education, public policy, and AI practice from a human-centered perspective.
In May 2026, it expanded by merging with the Stanford Data Science Institute into a single institute that retains the Stanford HAI name. This brings data science work under the “new HAI,” gives it stronger ties to industry and expands university computing capacity, and shows how major universities are reorganizing their institutional architecture around AI.
Stanford HAI produces the AI Index, one of the most influential reports for tracking the evolution of artificial intelligence. It gathers, verifies, and presents global data for governments, researchers, journalists, companies, and the general public. Since measurement is never a neutral task, by defining in quantitative terms what counts as a “notable model,” as well as variables for international comparison, investment, regulation, research, adoption, education, and computing capacity, Stanford HAI is part of the puzzle that is gradually establishing what we should understand as “progress” in the field of AI.
In its 2025 edition, the AI Index showed, for example, that nearly 90% of notable AI models in 2024 came from industry, compared with 60% in 2023. It also noted that U.S. institutions produced 40 notable models, compared with 15 from China and 3 from Europe. These data describe a trend and help consolidate an image of the world now grounded in empirical evidence: AI is increasingly dominated by companies, with the United States in the lead, China closing the gap separating it from the North American country, and Europe prioritizing regulation over the production of frontier models.
That is where Stanford HAI’s power lies: it does not directly set the rules of AI, but it helps build the language, metrics, and evidence that governments, companies, media outlets, and universities use to make decisions. This makes it a key actor in generating data on AI and translating that data into a diagnosis for the public agenda.
Basic facts
- Stanford HAI is based at Stanford University in California.
- In 2026, after the merger with Stanford Data Science, the institute came under the leadership of James Landay, a professor of computer science and specialist in human-computer interaction, accompanied by Russell Wald as Executive Director; Fei-Fei Li as Founding Director; and a group of Faculty Associate Directors from different fields, including Russ Altman, Emma Brunskill, Erik Brynjolfsson, Emmanuel Candès, Daniel E. Ho, Mykel Kochenderfer, Christopher Manning, Rob Reich, Ge Wang, and Amy Zegart, among others.
- The merger brought Marlowe, Stanford Data Science's high-performance computing cluster, into the HAI Institute.
- The Institute was founded in 2019 by Fei-Fei Li, John Etchemendy, Christopher Manning, and James Landay. After the 2026 reorganization, Fei-Fei Li also became special AI adviser to Stanford President Jonathan Levin, while Li and former Stanford President John Hennessy became co-chairs of the institute’s advisory council.
- One strong indicator of its scale: in May 2026, Stanford reported that the merger between HAI and Stanford Data Science combined HAI’s network of more than 400 scholars, its industrial affiliates program, and $60 million in cumulative grant funding.
- Its 2023 annual report recorded $20.8 million in total expenses and $6.8 million issued in research grants: HAI functions as an internal funder of interdisciplinary research. In April 2026, it announced $2.17 million in seed grants to 29 research teams, supporting projects in areas such as health, scaling, collaborative coding, and other interdisciplinary lines of work.
- HAI has an industry affiliates program and has brought in affiliates such as Itaú Unibanco, AXA, and Infosys.
- In January 2026, Stanford HAI announced a partnership with the Swiss National AI Institute, ETH Zurich, and EPFL to develop open-source foundation models guided by social values rather than commercial interests.
- HAI brings together several internal centers and labs. Among the most relevant for your work are the Stanford Digital Economy Lab, associated with Erik Brynjolfsson, and the Center for Research on Foundation Models, led by Percy Liang, which connects HAI to debates on transparency, evaluation, benchmarks, and foundation models.