from the field

This story was originally published on Subvert’s blog https://subvert.fm/blog/reflections-from-the-platform-cooperativism-conference-in-istanbul/ on January 21, 2026

For those of you who don’t know, Subvert is part of a small corner of the tech world called “Platform Cooperatives”. The idea is that there are many different kinds of cooperatives: worker cooperatives, producer cooperatives, retail cooperatives, etc. So what do we call online cooperatives, or internet infrastructure owned as a co-op? Platform co-ops.

This small subculture within tech often goes unnoticed in mainstream tech culture. It’s organized around simple questions. What if platforms were owned and controlled by the communities that rely on them instead of founders and investors? Could there be an Uber owned by drivers? An OnlyFans owned by sex workers? Or in the case of Subvert – could there be a Bandcamp owned by its artists and community?

Since 2015, this community has been convening at annual conferences organized by the Platform Cooperativism Consortium. This organization, led by Trebor Scholz, professor at The New School and longtime advocate for worker-owned digital platforms, has held a global focus, hosting these gatherings across the globe: in New York City, Hong Kong, Berlin, Rio de Janeiro, India, Kenya, and now Istanbul.

Subvert traveled to Istanbul and attended as part of a panel, with our trip generously sponsored by the Center for Cultural Innovation – the same non-profit organization that supported our early work, including a fellowship that helped us design and print our first zine run.

These conferences have always centered discussion on how digital platforms can be owned and governed by the people who use and depend on them. But this year felt different. As the world is consumed with AI hype, this conference turned its focus to the intersection of artificial intelligence and online cooperativism.

Why AI and co-ops?

Before getting into what was discussed, it’s worth naming where many of us in the Subvert community stand on AI: skeptical at best, hostile at worst.

Generative AI music is widely considered slop. Bandcamp recently introduced a policy which bans AI-generated music. And the news keeps surfacing examples of AI companies like Suno training models by scraping copyrighted work without consent or compensation. Within music, the conversations about AI are particularly hostile. So where do the conversations around AI and cooperatives intersect?

We came with an open mind to hear from academic researchers and practitioners alike what they see in this intersection.

The Solidarity Stack

Trebor Scholz opened the conference by introducing what he calls the “Solidarity Stack” – a framework he developed with researcher Morshed Mannan for thinking about cooperative ownership across the entire AI supply chain.

“The AI crisis won’t be solved by laws or better code alone. It will be solved by people—by you—organizing across layers, borders, and disciplines. Across the stack.” – Trebor Scholz

The core idea he presented is that AI isn’t something that just exists at the code level. AI systems are produced by many layers: rare earth mineral mining for chips, data centers, human labor, data collections, algorithms, models, and apps. What Trebor outlined is a vision where each of these steps can be run as cooperatives, rather than controlled by a small number of corporations.

The Solidarity Stack asks: what if each of these layers could be cooperatively owned? What if there were cooperative mines, community-owned data centers, worker-owned AI development shops, and democratically governed applications?

It’s an ambitious, broad vision. And the realization of this vision feels distant and hopeful – bordering on impossible. But Trebor’s framework provides a useful map for seeing where intervention might be possible, and where the cooperative movement is already experimenting.

The hidden labor of AI

The conference opened with a screening of In the Belly of AI, a documentary directed by Henri Poulain and co-written by sociologist Antonio Casilli. The film traces the invisible human labor that powers AI systems: the data annotators, content moderators, and ghost workers scattered across the Global South who train models for a few dollars a day.

The movie illustrated that none of these AI models exist without exploited human labor.

“It depends on who owns it”

It was interesting that at this event, there was less overt dismissal of AI than we commonly see within Subvert’s membership. These researchers and practitioners shared a view that the root problem of AI exists within its capitalist business models, and less so within the technology

Transkribus, an AI tool for transcribing handwritten historical documents, must be one of the most successful cooperative platforms today. The platform has over 300,000 registered users, has processed more than 100 million pages of historical documents, and is co-owned by 237 members from 35 countries. It even won the European Union’s Horizon Impact Award in 2020.

Transkribus got started within academic research projects funded by the European Commission to read handwritten texts. The research that created this tool became valuable and useful enough that people around the world wanted to use it, and a community and business formed around it. When the EU funding ended in 2019, they transitioned into a cooperative rather than selling to a larger company.

Melissa Terras, a professor at the University of Edinburgh and director of its Centre for Digital Scholarship, spoke about how the cooperative has operated for years, resisting acquisition offers and maintaining democratic governance.

Professor Terras had this to say about AI:

“AI is just lines of code. It depends on what we point it at and who owns it.”

Her argument is that the problem isn’t the technology – it’s the business models. Cooperative structures are alternatives to extractive business logic.

AI optimism: can it be a “labor commons”?

Marcelo Vieta, a professor at the University of Toronto who studies worker cooperatives and self-management, offered the most optimistic talk of the conference. He said,

“We built these systems. We can redirect them. Or slow them down and re-design them.”

Vieta suggested that AI can become part of a “labor commons” – allowing AI, data, and algorithms to be shared resources collectively governed by workers, instead of commodities extracted for profit. He pointed to The Drivers Cooperative in New York as an example: the tech is similar to Lyft and Uber, but the ownership and governance are entirely different.

His point wasn’t that we should seize existing AI infrastructure. It was that we can build alternatives from the ground up. Or as he put it:

“recoding architectures, data practices, and incentive structures around human scale and ecological limits.”

He asks:

Can AI be a redemptive technology?

However, it’s worth noting that when an audience member inquired whether any architects of the dominant AI systems were in the room, they weren’t. To me, some of the optimism was dampened by this feeling that a group of researchers and academics were raging against the machine halfway across the world from San Francisco, while the AI architects we were raging against were neither present nor aware that this conversation was happening.

Cooperatives in practice

I shared my panel with Stuart Fulton, who runs PescaData, a data platform used by small-scale fishing cooperatives in Mexico to share information with each other. PescaData is worlds away from the traditional Silicon Valley tech world, so it’s refreshing to see smaller scale platforms from Latin America be represented. I appreciate the curation of Trebor Scholz, who has continued to extend a global lens to these events, making sure not to create a US-centric echo chamber of discourse.

Hypha Worker Co-op, based in Toronto, presented RooLLM. Roo is the name they gave their own internal, self-hosted AI tool which helps the workers within their cooperative coordinate by making internal company data and insights more accessible, and even support their own governance. It’s a small but real example of what it might look like for a cooperative to run AI infrastructure on its own terms. No reliance on OpenAI or Google. And no replacing jobs. Just experimenting with ways to empower worker-owners within a co-op.

Felix Weth, CEO of Platform21, made the case for cooperatives to invest in building shared compute infrastructure. He argued that without access to processing power, cooperative AI will always be dependent on the very corporations it hopes to challenge. His call to action: build the world’s largest collectively owned compute cluster.

Research vs Practice

In a world of very well-placed tech cynicism, there is something nice about being in a room with optimistic and hopeful tech nerds. I first attended a Platform Cooperativism Conference in 2019, in New York City. Back then, there was an energy that I found to be contagious. The people there wanted to build cooperative replacements for Facebook, Google, everything. The optimism was real.

Six years later, some things have changed. It’s true that platform cooperatives have made progress, but the ambition of the early days has dampened. Transkribus shows that there are real successful examples in practice today. That’s real progress. Still, I couldn’t help but feel some disappointment by the imbalance of the discourse.

There weren’t enough practitioners centered in the discussions I heard. Researchers and academics far outnumbered people who had actually started or were running cooperatives. The conversations often felt like observing rather than participating. There were more people studying the possibilities of a movement rather than picking up a shovel.

This isn’t necessarily the fault of the conference organizers. The Platform Cooperativism Consortium has roots as an academic organization, after all. But if the cooperative movement is going to matter in the AI era, the discourse needs to be reclaimed by people in the arena.

I have always felt that the health and success of the platform cooperativism movement is contingent on moving it away from the world of dry academic PDFs and research – and bringing the ideas to life through real, successful, consumer tech applications. But the space is largely still firmly planted in academia.

Whenever I am at a platform cooperative event, I can’t help but compare it to traditional tech events like TechCrunch Disrupt, where founders, funders, and journalists are centered in conversation. Say what you want about the traditional tech world – it’s organized around people building things. And I think the cooperative movement needs more of that energy. More praxis to balance the theory.

Maybe the answer is for practitioners like Subvert to organize our own convenings, bringing together the people actually running cooperatives and creating space for the conversations we need to have. Less theory, and more practical examples of artists joining together to create their own worlds and systems.

Continued Work

Despite my critiques, there was something energizing about being in a room with people from thirty countries, all trying to figure out how to build something different.

One of the things that I loved was hearing genuine enthusiasm from many of the conference participants about what Subvert is doing. It was a reminder of the potential for impact we can have – through demonstrating our ability to be commercially successful against a legacy incumbent.

And it’s also a reminder that there is room for developing our own positioning on AI as it relates to our cooperative. This is an invitation for everyone reading this: we’re currently co-creating our own AI policy in the Subvert forum. If you haven’t weighed in yet, now’s a good time.

The questions raised in Istanbul about who owns and controls AI mirror Subvert’s questions about who owns and controls music platforms. These questions will only become louder. And the answers will be shaped by those who get in the arena and participate.