Why Stanford HAI Rules and Spanish Universities Are Nowhere

Today we're talking about the trail institutional laziness leaves when a decade sustains it. Stanford HAI publishes the AI Index. It's cited by the Financial Times, the New York Times, the Wall Street Journal, the Economist. In Spain there's no equivalent. Probably my ignorance keeps me from knowing some paper circulating through narrower channels —there must be serious researchers working in well-meaning departments with no megaphone. Well, I don't know. What I do know is that I lived through another era in the public university. Another era where people with real prestige came to give lectures, and where a certificate meant something different from today's theater. What happened to us along the way?

What Stanford HAI Really Is

Human-Centered AI Institute. Founded in March 2019 by Fei-Fei Li, John Etchemendy, Christopher Manning and James Landay. It's not a department or a faculty. It's a cross-cutting institute with its own funding, a multidisciplinary team and an independent editorial agenda. That word —editorial— is the key so often overlooked.

A university department produces papers and trains doctoral students. HAI does that and on top of it manufactures documents designed to enter the public debate: the AI Index, the policy briefs, the open seminars. The difference isn't academic, it's a matter of craft. It isn't academia talking to academia. It's academia talking to whoever decides in Brussels, in Washington and in the newsrooms that count.

The AI Index Report 2026 runs over four hundred pages. Nine chapters. An uninterrupted annual series. Each chart carries explicit methodology. Each figure carries a source. When the global press cites the figure of 53% adoption of generative AI in three years, it cites it because it can go to the PDF and check where it comes from.

The Spanish Landscape, Without Makeup

Spain has serious technical centers. The IIIA-CSIC, in Bellaterra, was constituted as an institute in 1994 out of an AI group that had been working since 1985 —yes, 1985— with solid lines in multi-agent systems, machine learning and reasoning. The Barcelona Supercomputing Center does front-line work in compute and models. The CNIO has powerful groups applying AI to oncology.

There are decent doctoral theses and small cathedrals with competent professors.

What there isn't is an editorial product.

According to its own website, the IIIA brings together more than 70 scientists and engineers, a couple of dozen of them doctoral students. There's no annual report with its own brand that the global press cites. There are papers at serious conferences —IJCAI, AAAI— and no 400-page document that serves as a cross-cutting reference. The consequence is direct: when a journalist at El País or El Mundo needs a figure on the state of the sector, they don't call the IIIA. They call OpenAI. Or they cite Stanford translated from the previous day's NYT.

This isn't a failing of the researchers. It's a structural failing of how the Spanish public institution is conceived. The university and the CSIC produce knowledge that stays inside the academic circuit. Getting out requires an editorial infrastructure —communications, design, translation, distribution— that no Spanish AI center has at the necessary level.

A Decision, Not a Budget

The immediate objection is the economic one. Stanford has an endowment, private funds, massive donations, six-figure salaries. The CSIC lives off the General State Budget. Comparing them would be unfair.

It's true. And it's a half-truth.

Because Stanford HAI isn't only money. It's a decision about where to spend the money there is. HAI was born of a strategic choice: to turn AI into the university's cross-cutting editorial product. In Spain, the strategic choice of many universities was the opposite —to turn AI into a commercial product. An online master's at 13,000 euros. A distance postgrad with moonlighting faculty. Courses for executives at 2,000 euros a weekend.

One choice produces the AI Index. The other produces lead-capture machines. The budget explains the scale. It doesn't explain the direction.

I knew another Spanish university. The one of the late nineties and early two-thousands. People with real weight came to give lectures —active researchers, not producers of corporate PowerPoints. The library made sense. The great halls weren't rented out for bootcamp certificates. Then came Bologna, came the mandatory paid master's degrees for access to regulated professions, came mediocrity legitimized by fees, came the capture.

Now any YouTuber with two years of using the OpenAI API sets up their academy and sells certificates. A group of pals runs a course and hands you a diploma. And the worst part isn't that they sell it —it's that in some places it passes. It passes in companies that hire, it passes in administrations that accredit, it passes even in the university itself, which offers master's programs in partnership with those bootcamps whose faculty doesn't have a single indexed publication. Ugh.

Meanwhile, out of Palo Alto comes, on schedule, the AI Index 2026. And the global press reads it.

The AI Index as a Piece of Power

It's worth looking at the AI Index not as a technical report but as a piece of power —sorry, "it's worth" sounds like a manual. Let me say it another way: the AI Index works as a technical report for whoever downloads it, and as a piece of power for whoever decides what gets counted and what doesn't. It orders the global conversation. It decides which metrics matter —number of frontier models per country, private investment, corporate adoption, reasoning exams. It decides which countries it counts and which it doesn't. It decides which companies have a presence and which appear in a chart's footnote.

That editorial power isn't neutral. Stanford wields it. And Stanford decides not to break out Spain as its own category in most of the charts. Spain goes inside "Europe," or it doesn't appear. It's not an attack on Spain. It's that Spain, from the AI Index's point of view, doesn't produce relevant data in the categories the index measures.

Being outside the dataset isn't contempt. It's a symptom. And the symptom says that the conversation about AI in Spain is imported, translated and discussed with a two-year lag because Spain itself has no productive apparatus of data on the sector to put it at the table.

Is there anyone at the CSIC, at the CRUE, at the Ministry of Science building an AI Index Spain? An annual report with explicit methodology, verifiable figures, broken-out European data, comparable charts, and the editorial will to get it into the FT and the Economist and El Confidencial and Politico Europe?

No. Not that I'm aware of.

The Conversation Is Imported in Translation

When a Spanish reader wants to understand the sector, they have no national site to go to with authority equivalent to Stanford HAI. They go to one of three places.

To a corporate press release from OpenAI or Google, translated by the technology section of a general newspaper. To an American blog whose translation arrives with a two-year lag and with the translator's bias. To a private master's program whose only interest in informing them is selling them the master's.

None of the three sources answers to the Spanish reader's interest. All three answer to another interest. And the reader who informs themselves this way doesn't build their own judgment —they build a reflex opinion about debates that were already closed in the United States when we opened them here.

This has a name and it's informational noise amplified by institutional capture. The public university doesn't compete for that space because it doesn't conceive of itself as an editorial agent. The private university competes, but its editorial product is the master's, not the report. The general media don't have the capacity to produce their own AI Index —that requires a team of thirty people with a mixed researcher/editor/designer profile working a whole year. No Spanish outlet has that. No Spanish public center attempts it.

The Precedent Almost No One Cites

Before Stanford HAI there was the AI Now Institute, founded in 2017 at New York University by Kate Crawford and Meredith Whittaker. Another cross-cutting institute. Another editorial model. Another annual production cited by the global press. Another agenda —more critical, less optimistic than HAI's.

The precedent matters because it shows the model is replicable. It doesn't require being Stanford. NYU did it. Oxford did it with the Future of Humanity Institute, which the university itself closed in April 2024. ETH Zürich has the AI Center. The University of Edinburgh has ELLIS Edinburgh. They all produce public documents, they all get the global press to cite them, they all create editorial traction.

Where's the equivalent in Madrid, in Barcelona, in Valencia, in Seville?

There isn't one. Not because there aren't good researchers —there are. But because no institution has made the strategic decision to leave the academic circuit and enter the global editorial circuit.

Three Ways Out, None Easy

If we accept the diagnosis, the reasonable ways out come down to three. None is entirely good. You have to choose.

Accept the status quo and recognize that the Spaniard reads Stanford in translation. Honest, decapitalizes the national debate.

Wait for the administration to fund a cross-cutting editorial institute with HAI ambitions. Realistic over the long term, slow, vulnerable to political change.

Build it from outside the university. Private platforms with academic rigor, serious blogs, technical newsletters in Spanish, popular-science books that don't infantilize, podcasts with active researchers. Build the editorial ecosystem the university didn't want to build.

That third way is the one that's happening, slowly, in some corners. I mention it without irony. Whoever today publishes in Spanish about AI with technical rigor and a critical voice is doing the work the public universities decided not to do.

Quick Definitions

• Stanford HAI: Human-Centered AI Institute, founded in March 2019. It publishes the annual AI Index Report and runs a policy and education program.
• AI Index Report: an annual Stanford HAI document that measures the state of AI with public methodology. 2026 edition: over 400 pages, nine chapters.

• IIIA-CSIC: Artificial Intelligence Research Institute of the CSIC, constituted as an institute in 1994 out of a group active since 1985, located in Bellaterra. Spain's main public center in the field.
• BSC: Barcelona Supercomputing Center. A public high-performance computing center, with groups in AI and language models.
• AI Now Institute: an institute founded in 2017 at NYU by Kate Crawford and Meredith Whittaker. Critical editorial production on AI.
• Endowment: a university's endowment fund. Stanford's reached some 40.8 billion dollars as of August 2025.

References

• Stanford HAI — AI Index Report 2026 (PDF, over 400 pages, nine chapters). The figure of 53% adoption of generative AI in three years comes from this edition. https://hai.stanford.edu/ai-index/2026-ai-index-report
• IIIA-CSIC — official site: institutional history (institute constituted in 1994, origin group since 1985) and staff composition ("more than 70 full-time researchers and engineers, including 24 doctoral students"). https://www.iiia.csic.es/es/sobre/mensaje-del-director/ ; https://www.iiia.csic.es/es/people/personal-y-estudiantes/ ; https://es.wikipedia.org/wiki/Instituto_de_Investigaci%C3%B3n_en_Inteligencia_Artificial
• Stanford Report, October 2025 — value of Stanford's investment portfolio and endowment (≈40.8 billion dollars as of August 2025). https://news.stanford.edu/stories/2025/10/report-investment-portfolio-value-endowment
• University of Oxford — closure of the Future of Humanity Institute in April 2024. https://en.wikipedia.org/wiki/Future_of_Humanity_Institute
• Acemoglu, D. & Johnson, S. — Power and Progress (PublicAffairs, 2023).
• Mazzucato, M. — The Entrepreneurial State (Anthem, 2013).
• AI Now Institute — annual reports 2017-2024.

También te interesa

• What the AI Index Report 2026 Says That the Spanish Press Doesn't Tell You
• Where AI Is Really Thought About, and It's Not on LinkedIn
• Why AI Master's Degrees in Spain Aren't Worth Your Money
• How the Press Is Using the AI Index 2026 Figures

En otros sitios

• ELLIS: European network of excellence in AI (cross-cutting institute model)
• Barcelona Supercomputing Center: Spanish public center for compute and models