Non-profit Organizations
Last reviewed
May 11, 2026
Sources
24 citations
Review status
Source-backed
Revision
v2 · 2,500 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
May 11, 2026
Sources
24 citations
Review status
Source-backed
Revision
v2 · 2,500 words
Add missing citations, update stale details, or suggest a clearer explanation.
Non-profit organizations active in artificial intelligence are mission-driven entities that conduct research, set standards, advocate for policy, and provide education without returning surplus revenue to shareholders. In the United States such groups are typically incorporated under section 501(c)(3) of the Internal Revenue Code; in Germany they take forms such as the registered association (eingetragener Verein, e.V.) or charitable limited liability company (gemeinnützige GmbH); in the United Kingdom they register with the Charity Commission or operate as company-limited-by-guarantee structures. Tax-exempt status requires a charitable, scientific, or educational purpose, and US groups must publish Form 990 filings.
The AI nonprofit sector grew rapidly after 2014 as researchers, philanthropists, and civil society leaders sought institutional homes for basic research, AI safety work, ethics inquiry, and open data efforts that did not fit inside corporate AI labs or single university departments. Distinct from for-profit firms such as Anthropic (a Delaware public-benefit corporation) and Hugging Face (a private company), AI nonprofits report to boards of directors, depend on grants and donations, and often release datasets and models under open licenses. Their work has shaped public debates including the 2023 Statement on AI Risk and the Bletchley Declaration signed at the UK AI Safety Summit.
For-profit AI companies face pressure to ship products, protect intellectual property, and generate revenue, which can constrain transparent research and long-horizon safety work. Nonprofit structures help in several ways:
The table below lists groups with publicly documented nonprofit status (US 501(c)(3) or 501(c)(6), Canadian not-for-profit corporation, German e.V. or gGmbH, UK charity, or equivalent).
| Organization | Founded | Country | Legal form | Primary focus |
|---|---|---|---|---|
| Allen Institute for AI (Ai2) | 2014 | US | 501(c)(3) | Open AI research, OLMo models |
| Vector Institute | 2017 | Canada | Not-for-profit | Machine learning research |
| Mila, Quebec AI Institute | 1993 (NPO 2018) | Canada | Not-for-profit | Deep learning research |
| EleutherAI Institute | 2023 | US | Nonprofit institute | Open language models |
| LAION | 2021 | Germany | Registered e.V. | Open multimodal datasets |
| MIRI | 2000 | US | 501(c)(3) | AI alignment research |
| Future of Life Institute | 2014 | US | 501(c)(3) | Existential risk, AI policy |
| Center for AI Safety | 2022 | US | 501(c)(3) | Technical AI safety |
| Centre for the Governance of AI | 2018 | UK | Independent nonprofit | AI governance research |
| METR | 2023 | US | 501(c)(3) | Frontier model evaluation |
| Redwood Research | 2021 | US | 501(c)(3) | AI alignment research |
| AI Now Institute | 2017 | US | 501(c)(3) | AI ethics and policy |
| Partnership on AI | 2016 | US | 501(c)(3) | Multistakeholder principles |
| AlgorithmWatch | 2017 (gGmbH) | Germany | Charitable limited company | Algorithmic accountability |
| CDT | 1994 | US | 501(c)(3) | Digital rights, AI policy |
| Electronic Frontier Foundation | 1990 | US | 501(c)(3) | Civil liberties, AI policy |
| CSER | 2012 | UK | University centre | Catastrophic risk research |
| Leverhulme CFI | 2016 | UK | University centre | AI and society research |
| CHAI | 2016 | US | University center | Beneficial AI research |
| DAIR | 2021 | US | Fiscally sponsored | Community-rooted research |
| AI4ALL | 2017 | US | 501(c)(3) | AI education and outreach |
| Code.org | 2013 | US | 501(c)(3) | CS and AI literacy |
| MLCommons | 2020 | US | 501(c)(6) consortium | MLPerf benchmarks |
| LF AI & Data | 2018 | International | Linux Foundation project | Open source AI tooling |
| Open Source Initiative | 1998 | US | 501(c)(3) | Open Source AI Definition |
| Mozilla Foundation | 2003 | US | 501(c)(3) | Public-interest AI |
The Allen Institute for AI, known as Ai2, was founded in Seattle in 2014 by Microsoft co-founder Paul Allen as a 501(c)(3) research institute. It has released the Semantic Scholar literature search engine and the OLMo family of fully open large language models with their training data and checkpoints.
The Vector Institute, established in Toronto in 2017 with a CAD $135 million commitment from Canadian governments and industry sponsors, is an independent not-for-profit specializing in machine learning and deep learning. Together with Mila (founded by Yoshua Bengio in Montreal, a nonprofit since 2018) and the Alberta Machine Intelligence Institute (Amii), Vector forms the network of three federally supported institutes under the Pan-Canadian AI Strategy.
LAION, the Large-scale Artificial Intelligence Open Network, was incorporated in Germany in 2021 as a registered association (e.V.). Led by Christoph Schuhmann, the volunteer group released the LAION-400M and LAION-5B image-text datasets used to train Stable Diffusion and other text-to-image models, along with OpenAssistant data.
EleutherAI began in July 2020 as a Discord collective coordinating open replication of GPT-3. In 2023 the group incorporated as the EleutherAI Institute, funded by Stability AI, Hugging Face, Canva, Nat Friedman, and Lambda Labs. Its outputs include the Pile training corpus and the Pythia open model suite.
The Machine Intelligence Research Institute (MIRI), founded in 2000 by Eliezer Yudkowsky with Brian and Sabine Atkins as the Singularity Institute for Artificial Intelligence, is a Berkeley 501(c)(3) focused on AI alignment; it renamed itself MIRI in 2013.
The Future of Life Institute (FLI), founded in March 2014 by MIT cosmologist Max Tegmark with Skype co-founder Jaan Tallinn and others, is a 501(c)(3) headquartered in Pennsylvania. FLI organized the 2017 Asilomar conference that produced the 23-point Asilomar AI Principles and published the March 2023 open letter calling for a pause on training systems more powerful than GPT-4.
The Center for AI Safety (CAIS), founded in 2022 by Dan Hendrycks and Oliver Zhang, is a San Francisco 501(c)(3) that operates compute clusters for safety research. In May 2023 CAIS published the one-sentence Statement on AI Risk: Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war. The statement was signed by hundreds of academics, lab executives, and public figures.
METR, founded as ARC Evals inside the Alignment Research Center by Beth Barnes, was spun off in December 2023 as an independent 501(c)(3) and renamed Model Evaluation and Threat Research. METR conducts pre-deployment evaluations of frontier models from OpenAI and Anthropic, contributing to system cards for GPT-4o and Claude. Redwood Research, a Berkeley 501(c)(3) founded in 2021, runs empirical alignment experiments and the Constellation coworking space.
The Centre for the Governance of AI (GovAI), founded in 2018 by Allan Dafoe within Oxford's Future of Humanity Institute, became an independent nonprofit in 2021 based in London under Ben Garfinkel. Apollo Research, an AI safety evaluations group fiscally sponsored by Rethink Priorities, has announced plans to transition into a public-benefit corporation.
The AI Now Institute, founded at New York University in 2017 by Kate Crawford and Meredith Whittaker, is an independent 501(c)(3) publishing annual reports on the social and labor implications of AI. Partnership on AI (PAI), launched in September 2016 by Amazon, Facebook, Google, IBM, and Microsoft and later joined by civil society partners, is a 501(c)(3) developing voluntary frameworks on synthetic media, labor, and inclusive research design.
AlgorithmWatch, founded in Berlin in late 2015, has operated as a gemeinnützige GmbH since February 2017 with offices in Berlin and Zurich, auditing automated decision-making systems and contributing evidence to the EU AI Act. The Center for Democracy & Technology (CDT), a Washington DC 501(c)(3) founded in 1994 by former EFF policy director Jerry Berman, runs an AI Governance Lab. The Electronic Frontier Foundation, a San Francisco 501(c)(3) founded in 1990, files amicus briefs on generative AI, copyright, and surveillance.
The Distributed AI Research Institute (DAIR), founded in December 2021 by Timnit Gebru after her departure from Google, is a fiscally sponsored project of the 501(c)(3) Code for Science & Society, with initial funding from the Ford, MacArthur, Open Society, and Rockefeller foundations.
The Centre for the Study of Existential Risk (CSER), a Cambridge centre founded in 2012 by Huw Price, Jaan Tallinn, and Lord Martin Rees, studies extinction-level threats from advanced technology. The Leverhulme Centre for the Future of Intelligence (CFI), based at Cambridge with spokes at Imperial College London and UC Berkeley, launched in 2016 with a GBP 10 million Leverhulme Trust grant. The Center for Human-Compatible AI (CHAI) at UC Berkeley was founded in 2016 by Stuart Russell. The Future of Humanity Institute (FHI) at Oxford, founded in 2005 by Nick Bostrom, closed on 16 April 2024 after the Faculty of Philosophy froze hiring and fundraising; its alumni populate many of the safety and governance nonprofits above.
AI4ALL, a 501(c)(3) founded in 2017 by Fei-Fei Li, Olga Russakovsky, and Rick Sommer (originating as a 2015 Stanford summer outreach program for high school girls), runs college pathways programs and open curricula. Code.org, the Seattle 501(c)(3) founded in 2013 by Hadi and Ali Partovi, launched the TeachAI coalition with Khan Academy, ISTE, and ETS to publish a global AI literacy framework and runs the annual Hour of AI campaign. Women in Machine Learning and Data Science (WiMLDS) is a 501(c)(3) supporting women and gender minorities through chapter-led workshops.
MLCommons, founded in 2020 as a 501(c)(6) industry consortium, maintains the MLPerf benchmarks; its AILuminate benchmark targets safety evaluation. LF AI & Data, an umbrella under the Linux Foundation, hosts projects including ONNX, Milvus, and OPEA; the related PyTorch Foundation was established in 2022 after Meta transferred stewardship of PyTorch. The Open Source Initiative, a California public-benefit 501(c)(3), released version 1.0 of the Open Source AI Definition in October 2024.
OpenAI, founded in December 2015 as the 501(c)(3) OpenAI, Inc., established OpenAI LP as a capped-profit subsidiary in 2019 to raise capital for training large models while the nonprofit board retained control. After years of restructuring debate, OpenAI announced in May 2025 that the nonprofit would continue to control the for-profit business. Mozilla Foundation, a 501(c)(3) established in 2003, has directed over USD 20 million in grants into public-interest AI through programs such as the Democracy x AI Incubator; the separately incorporated Mozilla.ai is a for-profit subsidiary.
| Resource | Producer | Type |
|---|---|---|
| OLMo, Dolma, Tulu | Allen Institute for AI | Open models and training data |
| Semantic Scholar | Allen Institute for AI | Scientific literature corpus |
| LAION-400M, LAION-5B | LAION | Image-text datasets |
| OpenAssistant | LAION and community | Conversational dataset |
| The Pile, Pythia, GPT-NeoX, GPT-J | EleutherAI | Open corpora and models |
| MLPerf, AILuminate | MLCommons | Benchmarks |
AI nonprofits rely on foundation grants (Open Philanthropy, MacArthur, Ford, Rockefeller, Kapor Center, Leverhulme Trust), government science funding (US National Science Foundation, UK Research and Innovation, the Canada CIFAR AI Chairs program), corporate gifts and sponsored memberships, and individual donations. Open Philanthropy, founded in 2017 by Holden Karnofsky, Dustin Moskovitz, and Cari Tuna and renamed Coefficient Giving in 2025, is the largest single funder of AI safety research and has directed over USD 4 billion in grants since its founding. The defunct FTX Future Fund, an effective altruism vehicle backed by Sam Bankman-Fried's firms, supported safety nonprofits in 2022 before the collapse of FTX reshaped donor patterns.
AI nonprofits face several persistent critiques. Many depend on data, compute, or staff from for-profit labs, creating conflicts of interest such as Partnership on AI's founding by large tech firms. The boundary between nonprofit and for-profit can blur: OpenAI's hybrid structure has been criticized as nonprofit branding for a commercial enterprise, and Apollo Research has moved toward public-benefit corporation status. The closure of FHI in 2024 and the collapse of the FTX Future Fund in 2022 illustrate fragility around hosting institutions and donor concentration. Existential-risk-focused safety nonprofits clustered around effective altruism have drawn criticism for prioritizing speculative future harms over documented current harms such as bias, labor displacement, and surveillance, a critique advanced by Timnit Gebru, AI Now, and others.
Despite these tensions, the AI nonprofit ecosystem has produced widely used open datasets, shaped public understanding of large-model risks, supplied technical evidence to governments at the Bletchley and Seoul AI Safety Summits, and trained researchers now working across academia, industry, and government.