Technological disruption? Co-operatives have been finding their own answers since 1844. Since then, they have put technology to work in farms, factories and, more recently, platform co-ops.
How do we carry those achievements, and the lessons from failures, into a story of change that recognises what is different about AI? What do we use, what do we refuse – and how do we prepare for the bubble to burst?
More than 160 participants from over 170 organisations in 40 countries are coming together in Bangkok for this year’s annual conference of the Platform Cooperativism Consortium. Their challenge is one the co-operative movement should recognise: how to turn scattered experiments into an economy people can govern together.
Solidarity AI takes place from 12–15 November, with Chulalongkorn University and PCC Thailand as key partners. Together with ICA Asia-Pacific, they are helping us pioneer and lead this work, with sponsorship from the Friedrich-Ebert-Stiftung. The conference builds on a conversation that Co-op News readers encountered last year. In his report from our Istanbul conference, Miles Hadfield examined what co-operatives were already doing with artificial intelligence, from managing shared data to developing alternatives to corporate platforms. (Main image: co-operators at the Istanbul event).
Our thinking did not stop in Istanbul. Looking closely at those projects, and others taking shape around the world, brought a difficult realisation. A co-operative can own a useful AI service while remaining dependent on infrastructure, models and labour arrangements over which its members have little control.
For readers familiar with co-operative wholesaling, this predicament has a history. Local societies built purchasing networks, factories and financial institutions because independence required co-operation at a larger scale. The co-operative commonwealth was an ambition to connect institutions capable of meeting everyday needs.
Bangkok offers another historical connection. Thailand’s first co-operative, Wat Chan in Phitsanulok, dates to 1916. Its purpose was to help farmers escape punishing debt. Today’s questions about dependence on technology suppliers belong within that longer struggle over who controls the conditions of economic life.
The stakes extend far beyond membership services or office productivity. AI now helps organise surveillance and warfare. Naza, the Guardian-produced documentary, examines Israeli military targeting systems in Gaza through testimony from military insiders. Its account forces questions about responsibility that cheerful discussions of innovation often evade.
The growing proximity between technology corporations and authoritarian governments makes these questions urgent. Amid mounting authoritarianism and neo-fascism in many countries, public institutions can become customers for systems that monitor, classify and exclude their own populations. Co-operative discussions of AI cannot leave this political setting outside the room.
Related: Co-op Congress: Looking to a mutual future for AI and digital
Technology also gives people formidable organising capacities. In Brazil, André Janones’s Lula’s Army WhatsApp channel helped coordinate supporters during the final three weeks of Lula’s election campaign. Millions joined a shared effort to circulate material and mobilise voters. Its precise electoral effect is difficult to isolate, but the organising capacity was unmistakable.
That mobilisation took place on infrastructure owned by Meta. Popular ingenuity and corporate dependence coexisted. Co-operatives know this tension whenever a democratically owned enterprise must organise its work through services whose owners answer to entirely different priorities.
Co-operation across the supply chain
Platform co-operatives have spent a decade asking what changes when workers own the platform that allocates their jobs. They, too, need federations and shared services. AI makes those dependencies especially extensive: even a modest experiment draws on computing infrastructure, datasets, models and human labour distributed across organisations and borders.
An assistant used by a co-operative may rely on a proprietary model, cloud servers, outsourced annotation and electricity generated elsewhere. Member ownership of the assistant does not establish democratic control over those relationships. Nor does using an open-source model automatically resolve them.
The human consequences can be devastating. Former moderator Fasica Berhane Gebrekidan’s investigation for the Data Workers’ Inquiry documents trauma, addiction and psychological distress among colleagues who screened disturbing material in Nairobi. Content moderation and AI data work are distinct activities, but their labour supply chains overlap, and both can conceal serious harm behind subcontracting.
This is why the seven co-operative principles, indispensable as they are, may be insufficient for governing an AI supply chain. A co-operative can practise member democracy while purchasing services built through exploitation. Concern for community must reach people who will never appear on its membership register.
Fairwork’s AI supply-chain framework offers a useful complement. It examines pay, conditions, contracts, management and representation, using evidence to assess labour practices and support improvements. Its evaluation and certification work gives co-operatives practical questions to ask suppliers, alongside standards against which to judge their own commitments.
A shared purchasing agreement, for example, could require suppliers to disclose subcontractors and demonstrate how workers can raise grievances safely.
Related: Global Innovation Coop Summit highlights co-ops
In Bangkok, those questions will bring co-operative practitioners into conversation with people beyond the movement. The programme includes African tech-worker organiser Kauna Malgwi, IT for Change’s Anita Gurumurthy, worker-rights campaigner James Farrar, philosopher Soraj Hongladarom and Transkribus scholar Melissa Terras.
Their contributions connect workplace organising, public policy, feminist analysis and Buddhist ethics to co-operative practice. No single organisational form possesses all the knowledge or power required. Municipalities, universities, unions and social movements belong in the work from the beginning.
That broader alliance changes what counts as a successful experiment. A convincing demonstration must address who maintains it, who pays when it fails and who can challenge its decisions. Technical performance matters, but a co-operative service also needs rules for resolving disputes and protecting shared assets when members leave.
The commons is central to this approach. Participants can contribute authorised knowledge, software and resources under rules they shape together. Sharing need not mean surrendering control or making every dataset public. A community may have good reasons to restrict access, especially where cultural knowledge or personal information is involved.
Three experiments worth connecting
Transkribus shows how a clearly defined purpose can sustain co-operative AI. Developed by READ-COOP, it helps archives, researchers and others turn handwritten and printed historical records into searchable text. Its co-operative structure gives participating institutions a way to help shape a service they depend on.
The significance reaches beyond transcription. Archives hold knowledge that commercial training datasets may overlook, including records of workers’ organisations and co-operative movements. Institutions could collaborate on making those collections usable while deciding what they authorise others to access. A knowledge commons requires stewardship, rather than indiscriminate extraction.
In Argentina, technology co-operatives Farox, Cambá and Nayra are developing GenIA, an open-source approach to deploying AI with greater control over organisational data and infrastructure. Hosting arrangements, supplier dependence and the terms of access become decisions organisations can examine, rather than conditions accepted by default.
Here, co-operation can extend to the people maintaining the system. Several enterprises could jointly employ developers or administrators, pool security expertise and share computing costs. A smaller organisation gains capabilities it could not afford alone, while retaining a voice in how those capabilities develop.
Thailand’s Origins Co-op brings the discussion back to land and livelihoods. Its regenerative rubber-farming network connects producers and agricultural information. The relevant question is what collective control of farm data enables farmers to decide, and how technology supports the relationships on which regeneration depends.
Origins need not become an AI laboratory to matter here. It illustrates why data governance should precede decisions about models. Farmers should help determine which information is collected, who can use it and whether an additional technology serves their purposes.
These projects occupy different parts of what we call the solidarity stack. Infrastructure underpins data and models, while labour runs through every layer. Economic democracy means giving workers and affected communities power across these connections, including over the benefits and ecological burdens they produce.
Choosing where to intervene
A practical starting point is to identify needs that dominant providers serve poorly. Could Basque co-operatives, public institutions and language communities commission a language model under shared governance? Could Kerala’s co-operative movement work with universities and Malayalam speakers on a system responsive to local knowledge?
These are proposals for investigation, not announcements of completed projects. They would need evidence of demand, appropriate data permissions, technical competence and sustained funding. Their value would depend on serving a community, rather than matching the scale of the largest commercial models.
Solidarity AI can draw on open-source and proprietary models. The test is what control communities retain, what dependencies they accept and whether they can change course. Sometimes a smaller model will suffice. Sometimes a simpler tool will do the job without AI.
Regulation remains necessary. Rights need enforceable protections, and corporate power needs public limits. But enforcement depends on political capacity and governments willing to act. Workers cannot suspend their efforts to build a democratic economy until those conditions improve.
The co-operative contribution should therefore include dependable incomes and useful services, as well as democratic procedures. People need tangible reasons to trust institutions that ask for their participation. Affordable provision and meaningful influence over work can strengthen democracy where insecurity otherwise makes authoritarian promises attractive.
Experiments can make alternatives concrete while organising creates pressure for stronger rules. Public procurement can provide reliable customers. Co-operative finance can support shared assets. Unions can bargain over deployment and protect the right to refuse it.
No organisation has to reproduce the entire stack within its own walls. A federation might arrange computing, a university provide research capacity and a municipality commit to purchasing a service. The arrangement needs transparent responsibilities and a credible exit route, so collaboration strengthens autonomy rather than quietly replacing one dependency with another.
Bangkok’s task is to turn encounters into commitments. Planned outputs include the Bangkok Theses on Democratic Alternatives to Big Tech AI, a shared glossary of Solidarity AI, a map of working alternatives, and recorded voices from across the region. Documents emerging from this work will be directed to the United Nations, bringing co-operative and community perspectives into international debates on AI.
Every session has a shared document for community notes, open to participants in the room and online. These will record lessons, disagreements and concrete next steps for that session. Each day’s notes will also feed into the collectively written Bangkok Theses, to be read aloud at the closing and published afterwards.
After Bangkok, we will follow up on those next steps, connect collaborators and build a network of action. The test is what participants can do together once they get home.
