Published to coincide with the launch of Stanford HAI’s 2026 AI Index — April 13, 2026
When Stanford’s Institute for Human-Centered AI (HAI) needed to answer one of the most consequential questions in technology today — where in the world is frontier AI talent actually being built, and where is it moving? — they turned to Zeki.
We are proud to share that Zeki is a data collaborator on the 2026 Stanford HAI AI Index, the most authoritative annual benchmark of progress in artificial intelligence. Our data underpins the Index’s analysis of AI authors and inventors in Chapter 1 (Research and Development) and informs Chapter 8’s framework on talent sovereignty — a new dimension of national AI capability that policymakers, investors, and research leaders are only beginning to grapple with.
Why this matters
The AI Index is read by heads of state, central banks, Fortune 500 boards, and the labs building frontier models. When it cites a number, that number shapes decisions. The fact that Stanford selected Zeki to provide the underlying view of the global AI talent pool reflects something we’ve spent years building: a dataset that doesn’t exist anywhere else.
Most “talent data” on the market is really job-posting data, LinkedIn scrapes, or self-reported skills. That tells you who calls themselves an AI professional. It does not tell you who is actually pushing the science forward.
Zeki does something fundamentally different. We identify the 658,000 individuals worldwide (outside China) with a proven track record of producing genuinely new AI — people whose contributions appear in published research, open data depositories, and new models. These are the scientists and engineers who create intellectual property rather than apply someone else’s. They are the rarest and most valuable people in the technology economy, and they are extraordinarily hard to find at scale.
What the AI Index found using Zeki data
The Index’s findings, drawn from Zeki’s longitudinal dataset spanning 2010 to 2025 across 21 countries, tell a story that the headlines have missed:
- The United States remains the dominant home of frontier AI talent, with 220,520 identified authors and inventors in 2025 — but its net inflow advantage has collapsed from a peak of 324.6 in 2022 to just 26.0 in 2025.
- Switzerland and Singapore lead the world on a per-capita basis, with roughly 110 frontier AI researchers per 100,000 inhabitants — far ahead of any large economy.
- India has become a major source of frontier talent (50,460 in 2025) and its best minds are increasingly staying at home, losing only 16.9 net researchers to overseas opportunities in 2025 compared to hundreds in past years.
- Despite a decade and a half of industry growth, the gender ratio among frontier AI talent has barely moved since 2010. In Brazil, South Korea, and Japan, more than 80% of identified frontier talent is male. No country is close to parity.
- National specialisations are sharpening: South Korea concentrates in hardware and VLSI (consistent with its semiconductor role), Brazil in software engineering, Saudi Arabia in security and cryptography.
These are not opinions. They are observations grounded in career histories, research output, and patent activity tracked person by person, year by year, across the entire dataset.
What makes Zeki’s data rare
Three things separate Zeki’s work from anything else on the market.
The unit of analysis is the individual contributor, not the job posting. Every person in the Zeki dataset has been identified through their actual scientific output. That single methodological choice changes everything downstream — it is why the AI Index can talk about mobility, net flows, and specialisation shifts with confidence, rather than guessing from vacancy data.
The taxonomy is built from the science itself. Zeki has curated the areas of AI innovation across software, hardware, and compute through systematic analysis of the research literature. When we say someone specialises in computer vision or in VLSI design, that classification reflects what they actually publish and build — not a self-reported headline on a profile page.
The data is collected compliantly and transparently. Zeki works only with publicly available, business-related information that individuals or organisations have themselves chosen to publish. We never aggregate data from behind login walls, and any third-party data we license is vetted by external legal experts for regulatory compliance. For an institution like Stanford HAI, whose credibility depends on the integrity of every input, this is non-negotiable — and it is why we were trusted with this work.
Smoothing methodology, 12-month moving averages to handle profile-update lags, probabilistic gender inference enhanced by country of residence, longitudinal panels that follow careers rather than snapshots — these are the unglamorous engineering decisions that make a dataset trustworthy at the level the AI Index demands.
Talent sovereignty: a new lens for a new era
Perhaps the most significant contribution of Zeki’s data to this year’s Index is enabling Stanford’s framework on talent sovereignty — the idea that a country’s AI future depends not only on compute, capital, and policy, but on its ability to develop, attract, and retain the humans who actually build the technology. For the first time, governments have a defensible empirical picture of where they stand on that dimension. That picture is built on Zeki data.
This is the kind of work Zeki exists to make possible. If you are shaping AI strategy, policy, or investment decisions, you need more than hot takes and headcount estimates. You need to know who the people are, where they are, what they work on, and where they are going next.
That is what we do. And we are honoured that Stanford HAI agreed.
Read the full 2026 AI Index report here.
Want to talk to Zeki about how this data can support your organisation’s strategy, policy, or investment work? Get in touch with our team.
Zeki is a data collaborator on the 2026 Stanford HAI AI Index. #AIIndex2026 @StanfordHAI