Distributed AI Research Institute

The Distributed AI Research Institute (DAIR) is an independent, community-rooted AI research institute founded by Timnit Gebru on 2 December 2021 — the one-year anniversary of her departure from Google's Ethical AI team — and operated as a fiscally sponsored project of Code for Science & Society. DAIR is headquartered in Oakland, California, with a team distributed across 18+ countries, 7 time zones, and 20+ languages — a structurally pan-African-diaspora team spanning Ethiopia, Ghana, South Africa, Kenya, Europe, Australia, and the United States. The institute's working position is that AI research, development, and deployment should not be defined inside the incentive structures of Big Tech, that the people most affected by AI systems should lead the research about those systems, and that "AI is not inevitable, its harms are preventable, and when its production and deployment include diverse perspectives and deliberate processes it can be beneficial" — Gebru's own framing of the institute's founding theory of change.

Founding

DAIR was announced on 2 December 2021 — exactly one year after Gebru's contested 2 December 2020 exit from Google, where she had co-led the Ethical AI team — with an initial $3.7 million in founding funding from the Ford Foundation, the John D. and Catherine T. MacArthur Foundation, the Kapor Center, and the Open Society Foundations. Gebru's launch framing — "AI needs to be brought back down to earth. It has been elevated to a superhuman level that leads us to believe it is both inevitable and beyond our control" — positions DAIR as a corrective both to the discourse of AI inevitability and to the institutional structures that produce that discourse: research labs whose budgets, publication processes, and personnel pipelines run inside the same companies whose products they are studying. The institute was announced simultaneously as a Code for Science & Society fiscally sponsored project, making CS&S the organisational home of DAIR's work rather than DAIR itself being incorporated as an independent 501(c)(3) — a deliberate choice that allowed the team to launch in operational form without the overhead of building back-office infrastructure from scratch.

Structure and team

DAIR is led by Timnit Gebru as Founder and Executive Director, Alex Hanna as Director of Research (who joined from Google's Ethical AI team in February 2023), and Mike Medow as Director of Operations. The named research team includes Dylan Baker as Lead Research Engineer (who also transitioned from Google's Ethical AI team in February 2023), Milagros Miceli as Research Lead, and researchers Kathleen Siminyu, Adrienne Williams, Meron Estefanos, and Raesetje Sefala — Sefala having been the first research fellow announced in December 2021 for her work on geographical and economic segregation in South Africa using satellite imagery. Esra'a Al Shafei leads the Surveillance Watch programme. The fellows roster names Adio Adet-Dinika, Krystal Kauffman, Nyalleng Moorosi, and Asmelash Teka Hadgu as research fellows. The board is short — Safiya Noble (author of Algorithms of Oppression and a long-standing scholar of algorithmic discrimination) and Ciira wa Maina, named together as founding advisors in the 2 December 2021 fiscal-sponsorship announcement and now seated as board members. The faculty-affiliate layer adds Ellen Berrey, Elaine O. Nsoesie, Zeerak Talat, and Emily M. Bender — the linguist with whom Gebru and others co-authored the 2021 "Stochastic Parrots" paper whose internal review at Google preceded Gebru's departure.

Research approach

DAIR's three working principles — cutting through AI hype by demonstrating what AI can realistically do and where its limits are; grounding research in community expertise and local needs; and cultivating alternative technological futures through what the team calls imaginative discipline — translate into a deliberately slow, partner-led research process structured around the proposition that the people most affected by AI systems should lead the research about those systems. The institute's public framing treats researcher wellbeing as part of the methodology rather than as workplace policy, with explicit working-time commitments and family support designed against the exploitative working norms Gebru identifies in adjacent industry and academic AI research. The flagship 2021-launch research projects illustrated the approach: Sefala's spatial-apartheid analysis using satellite imagery to document the persistence of township-suburb segregation in post-apartheid South Africa, and Ernest Mwebaze's cassava-disease diagnostic work with smallholder farmers in Sub-Saharan Africa — both projects deliberately chosen to make the case that the questions worth asking with AI methods are not the questions Big Tech research labs prioritise.

Data workers and the Data Workers' Inquiry

DAIR's most corpus-load-bearing programme is the Data Workers' Inquiry — a "global, radically participatory research initiative spanning nine countries across five continents" where nineteen data workers themselves serve as community researchers documenting the labour conditions of the African, South American, European, and Middle Eastern annotators, content moderators, and chatbot-impersonation workers whose hand-classified examples underwrite the training of commercial AI systems. The initiative developed Workers' Inquiry as a Research Methodology (WIRM), and its outputs cover Sama / Cloud Factory / Remotasks content-moderation work in Kenya, refugee data annotation in Syria and Lebanon, platform-mediated data work in Brazil, and content-moderation and chatbot-impersonation work in France, Germany, Spain, and the UK — published as zines, documentaries, comics, essays, podcasts, and animations rather than only academic papers. DAIR is the initiative's primary funder, with the Weizenbaum Institute and Technische Universität Berlin as the named institutional partners.

The corpus's most direct line from DAIR into grassroots organising runs through this programme. The Data Labellers Association — Kenya's first worker-led association for data annotators, launched in Nairobi on 13 February 2025 — sits as one of the Data Workers' Inquiry's named projects, with DAIR identified in the DLA's own public page and in Computer Weekly's launch coverage as one of the international supporters that anchored the launch alongside AI Now, Turkopticon, the German Essen content moderators, and Foxglove. The DLA's public page is hosted on the Data Workers' Inquiry / DAIR infrastructure, making DAIR not just an external supporter but the load-bearing institutional home from which the DLA's first months of public organising operated.

Funding and posture

DAIR's funder roster — the Ford Foundation, the MacArthur Foundation, the Kapor Center, and the Open Society Foundations at founding — has been extended through subsequent renewals; the MacArthur Foundation in 2025 committed an $800,000 general-operating-support grant over 1 year and 10 months under its Technology in the Public Interest programme area, naming DAIR's three core areas as "using data to empower marginalized communities, developing ethical frameworks for AI research, and envisioning inclusive technological futures." The funder mix — foundation philanthropy with no corporate AI-company support — is itself part of the institute's working theory of change: DAIR's public framing treats independence from the entities whose products the institute studies as a structural prerequisite for community-rooted research rather than as a stylistic preference.

Posture in the movement

DAIR's distinctive contribution to the make-AI-good movement is structural. The corpus's existing research-and-advocacy entries — the Algorithmic Justice League (Boston / US, founded 2016 by Joy Buolamwini), the Ada Lovelace Institute (London, UK), and AlgorithmWatch (Berlin, Germany) — share an algorithmic-accountability research posture with DAIR but operate inside Northern Hemisphere institutional geographies and Northern Hemisphere staff distributions. DAIR is the corpus's first independent AI research entity built around an African-diaspora team distribution and around explicit movement-organising partnership with worker-led bodies in the AI supply chain. The institute's relation to Buolamwini and AJL runs through Gebru's own research record — Gebru was Buolamwini's co-author on the 2018 Gender Shades paper that seeded the algorithmic-accountability field — but the institutional relationship between DAIR and the African content-moderation / data-labelling cluster (the African Content Moderators Union, the Data Labellers Association, Foxglove, and the Foxglove–Nzili & Sumbi Kenya docket) is the more programmatically load-bearing link, with the Data Workers' Inquiry as the named convening infrastructure. The working theory of change is that AI's labour and political-economy dimensions are not separable from its technical and policy dimensions, and that civil-society research infrastructure has to be built at the geographic and demographic scale of the workforce whose labour underwrites the systems being studied.