Open-access learning, multilingual tutors, tools for neurodivergent learners, and the indigenous-knowledge stewardship that the dominant ed-tech narrative leaves out. AI is the most consequential teaching technology of the century. The catalog's job is to surface the work that uses it for emancipation rather than enclosure.
The dominant path for AI in education runs toward proprietary tutoring products marketed to wealthy school districts, opaque student data flows, and the quiet replacement of teachers with retention-optimised chat interfaces. The catalog's job here: surface the open alternatives, the multilingual work being done outside the English-language gravity well, and the tooling for the learners — neurodivergent, low-vision, deaf, low-bandwidth — the mainstream ed-tech market consistently skips.
Non-proprietary tutoring infrastructure and open courseware. The institutions and projects that have spent decades building learning at scale on something other than rent extraction.
The largest free-to-learner education platform on the open web. Khanmigo is its AI tutoring layer, available free to many school districts via partnership programs. The closest the field has to a serious nonprofit attempt at AI-native tutoring at scale — with student-data practices that don't read as a hostage situation.
Two decades of MIT course material released openly to the public — lecture notes, problem sets, video, exams. The single most important precedent for treating elite university content as a public good rather than a credentialing moat. Still updated, still free, still the reference implementation.
University-grade open courseware with partners including MIT, Harvard, Berkeley, and dozens of others. Audit tracks remain free; verified certificates monetise the platform without locking the underlying learning behind a paywall. A pragmatic compromise that has held up.
Wikimedia Foundation + Wiki Education
The Wikimedia Foundation stewards Wikipedia and its sister projects — the largest open-knowledge commons ever built. Wiki Education brings that infrastructure into university classrooms, training students to contribute peer-reviewed knowledge to the commons rather than essays that get filed and forgotten.
Free, ad-free early learning for under-eights, with literacy, math, social-emotional, and creative content from a nonprofit operator. The negative-space proof that early-childhood digital learning doesn't have to mean engagement-hacked attention extraction.
AI access outside the English-language gravity well. Speech, translation, and NLP for the languages the dominant model labs underserve — built by the communities that speak them or by long-running open-data efforts.
Crowdsourced multi-language speech dataset released under CC0. Coverage is uneven, but where it exists Common Voice is one of very few large-scale speech corpora available without restrictive licensing — and the contribution pipeline for adding under-represented languages is open to the communities themselves.
An African research-and-product lab building voice, translation, and NLP for African languages. Vulavula and InkubaLM target the long tail of languages that the major model labs route through English by default. Africa-based team, Africa-first roadmap.
No Language Left Behind (NLLB)
Meta's open-weight translation models targeting 200+ languages, with model weights, evaluation data, and research releases. Imperfect — many language communities have valid concerns about how their languages were sourced — but currently the largest openly available substrate for translation work outside the dominant five.
A grassroots NLP research community building machine translation, named-entity recognition, and speech infrastructure for African languages. Distributed, volunteer-led, with strong norms around language-community consent and credit. The model for what self-determined low-resource NLP looks like.
Mozilla Foundation language work
Beyond Common Voice, the Mozilla Foundation runs responsible-AI grants and fellowships that have funded language and accessibility tooling for years. Worth tracking as a recurring source of small-but-meaningful grants for community language projects.
Sovereignty over data and pedagogy. The infrastructure that lets indigenous and community-led knowledge be shared on the community's terms — not flattened into another scrapeable dataset.
Traditional Knowledge (TK) and Biocultural (BC) Labels — a metadata standard letting indigenous communities assert provenance, consent terms, and conditions of use on their cultural materials wherever those materials surface. A practical, deployable answer to "how do you mark indigenous data so it isn't silently re-extracted."
An indigenous-led, open-source content management system for digital heritage — built so cultural materials can be shared online with the access protocols (community, family, gender, ceremonial) that the originating community sets, rather than the defaults of mainstream archive software.
A campaign-and-research organisation working to centre marginalised knowledges — women, indigenous peoples, the global South — on the internet's open-knowledge platforms. Focus on Wikipedia and the broader commons, with research that consistently surfaces what is missing.
A working group of indigenous researchers, technologists, and knowledge holders developing position papers and protocols on indigenous data and AI. The closest the field has to a published frame for what indigenous-aligned AI development would actually look like in practice.
First Nations Information Governance Centre
FNIGC stewards the OCAP principles — Ownership, Control, Access, Possession — the foundational framework for First Nations data governance in Canada. The reference standard cited in most indigenous-data-sovereignty conversations across the Anglosphere.
Tooling the mainstream ed-tech market consistently skips. Visual, auditory, speech, and cognitive accessibility — treated as a primary use case rather than a checklist item bolted on at launch.
A long-running visual-interpretation service for blind and low-vision users, originally built on volunteer video calls and now offering an OpenAI-powered "Be My AI" mode that describes images and scenes on demand. One of the more credible examples of frontier-model integration delivering immediate access value rather than vapour.
Microsoft Seeing AI & Immersive Reader
Two of the most widely deployed pieces of free assistive AI from Microsoft's accessibility programme. Seeing AI describes the world for B/LV users via phone camera; Immersive Reader provides reading support across products in classrooms and workplaces, including for dyslexia and ADHD users. Free, no commercial lock-in.
Speech recognition trained for users whose speech is non-standard — cerebral palsy, ALS, stroke, traumatic brain injury, Down syndrome. The default speech-to-text products fail badly for these users; Voiceitt is one of the few products attacking that failure mode head-on rather than waiting for the dominant models to incidentally fix it.
Texthelp Read&Write
A reading-and-writing support toolkit deployed widely in schools and workplaces for dyslexia, ADHD, and EAL learners. Long-running product with a track record of treating accessibility as the core feature rather than a side mode — predating the current AI hype cycle.
Accessibility consultancies serving education
A growing layer of disability-led accessibility consultancies (Ablr, Fable, and others) work with ed-tech vendors and institutions to bring real users with disabilities into testing and procurement decisions. Listed as a category because the named players shift; the function of disability-led accessibility QA in the procurement chain is the persistent piece.
Infrastructure for the people doing the teaching. Open curriculum, evaluation, AI-literacy resources, and the educator organisations that have a track record on the policy questions before "AI in classrooms" was a beat.
An open educational resource library: free, openly licensed K-12 lesson plans, units, and materials contributed by educators. One of the longer-running OER projects, with a catalogue that maps to standards rather than to a single vendor's curriculum.
A nonprofit producing freely licensed health-education content — HIV/AIDS, concussion (CrashCourse), and adjacent public-health topics — localised across dozens of languages and used by hundreds of millions of learners. A working example of open educational content as global public health infrastructure.
Common Sense's education arm publishes ed-tech reviews, AI-literacy curricula, and student data privacy ratings used by school districts to make purchasing decisions. The closest the K-12 sector has to an independent consumer-reports layer for ed-tech and classroom AI.
Independent, educator-led reviews of K-12 instructional materials. A nonprofit operating on the principle that districts deserve honest, comparable reviews of curricula before committing to multi-year adoptions — rather than only the marketing the publishers send.
The International Society for Technology in Education. Long-running educator membership organisation with the most widely adopted technology standards for students, educators, and school leaders, and a growing body of guidance on AI in classrooms grounded in classroom practice rather than vendor whitepapers.
Capital aligned with access over enclosure. The funder ecosystem most realistic as multi-year support for projects in this domain — with the caveat that funder priorities shift, and inclusion here is observation, not endorsement.
The largest single education funder in the world, with US K-12 programs (math instruction, postsecondary success, learning analytics) and global education programs across South Asia and sub-Saharan Africa. AI in education is now a stated strategic priority area.
Carnegie Corporation of New York
A century-old education funder. Programs span teacher preparation, immigrant integration, and democracy — with consistent support for the institutional infrastructure of public education rather than buy-in to whichever vendor narrative is on offer in a given decade.
A major K-12 funder with an emphasis on charter schools, school choice, and education-reform networks. Listed for completeness because the funder ecosystem is what it is — not because every project in this domain will find Walton's priorities aligned with their own.
A research funder backing AI for science, AI safety, and AI in learning — with a willingness to fund infrastructure (datasets, evaluation, open tooling) rather than only publication-track research. Worth tracking for groups working at the AI-and-education research frontier.
Hewlett Foundation Open Education
Hewlett's Open Education program is the longest-running serious funder of open educational resources in the United States. A meaningful share of the OER substrate today exists because of Hewlett's multi-decade commitment to the field.
A funder explicitly oriented toward the social impact of technology — civic infrastructure, the future of work, learning. Patient capital with a research-and-practice posture, working at the edge of where ed-tech, civic tech, and labour intersect.
This page is a sample of the Education & Knowledge domain. Inclusion is editorial — not a commercial endorsement, not a vetting of every claim a project makes, and not a recommendation that any single tool fits every learner's situation. Foundations listed here are not partners; they're documented as the funder ecosystem most aligned with this work, public information, no relationship implied. If you build in this space and we missed you, submit your project below.
Beyond the catalog-wide exclusions, Education & Knowledge has its own line. Tools that wear the language of personalisation, equity, or rigour while operating as surveillance, punishment, or rights-laundering infrastructure are not the work. Naming the line is part of the work.
If your work fits the Education & Knowledge domain — and is honest about which line it sits on — we want it in the catalog. Submissions are reviewed editorially before inclusion. We may reach out for clarifying details.
Pulled live from the AI for Planet catalog — reviewed editorially, with maturity and evidence flagged. Inclusion is not endorsement; see the curatorial note below. Submissions for the Education & Knowledge domain open Q3 2026.