Melissa Terras on why cooperatives are the future of AI

Transkribus is run not by a company, but by a cooperative: READ-COOP. And according to board member Melissa Terras, it might be the only AI cooperative in existence.

As Professor of Digital Humanities at the University of Edinburgh and READ-COOP's Scholarly Director, she’s spent the past year and a half assessing different governance models in the AI sector for an upcoming research paper. And despite her in-depth research, she has yet to come across another AI organisation with a cooperative structure similar to READ-COOP. 

We talked to Melissa to hear more about her views on AI governance and about how cooperatives could present a brighter future for the technology.

The Transkribus funding dilemma

Melissa Terras has been part of the Transkribus from the very beginning. With a PhD in handwriting recognition, she first got involved in the project back in 2012, when the technology was still in its infancy and funded by a series of EU research grants. This money allowed researchers to create a prototype of Transkribus, that was successfully tested by institutions throughout Europe. But as the funding came to end, Melissa and the other project partners had a difficult decision about how to finance the platform further.

"There was a need from the community, which was "Please find us a plug-in-and-play version of handwriting recognition technology which we can use without having to figure it all out ourselves"," Melissa explained. "We wanted to fulfil this need. However, we wanted to do it in a way that respected the community across the library and archive sector."

Melissa (second from right) was one of the founding members of READ-COOP. © Transkribus

Beyond big tech: Exploring business models

Until Transkribus, most AI software for the culture heritage sector fell into one of two categories. "On the one hand, there's the open access and the open scholarship community, who have produced a whole bunch of tools. If you know what you're doing with computing, you can run your own code on your own servers and own that whole process. But a lot of libraries and archives simply don't have that skill set."

"At the other end of the spectrum, we have Open AI, Google, Microsoft, all of whom have some kind of image-to-text tool now. But it's not specific for historical text and it's not being trained up enough."

Melissa and her colleagues didn't feel that either of these models were right for Transkribus. "We've got this nice balance of setting up something which responds to the need from the community without needing to extract horrendous amounts of money from them. It's something that we can build together."

READ-COOP now has over 250 members, including the University of Cambridge, the National Archives of Finland, and the Rijksmuseum in Amsterdam. © Transkribus

Creating a cooperative

After much discussion about possible business models, it was decided to found a European cooperative society, a type of cooperative that is designed to operate across several EU countries. "Günter [Mühlberger, one of the founders of Transkribus] very much had the vision for setting up a cooperative," Melissa explained. "It would be commercial—we have to make a profit so that we can keep the lights on. But we wanted to set it up in a slightly different way that respected the community across the library and archive sector."

The cooperative structure lends itself to respectful collaboration. Instead of being owned and managed by shareholders and investors, READ-COOP is managed by its members. Anyone can become a member of the cooperative, giving you the chance to vote on a wide range of issues affecting the organisation at the cooperative's Annual General Meeting. There are also Monthly Members' Meetings, a Slack channel where members can talk to each other and the Transkribus team, and sometimes additional one-off meetings for urgent issues.

"[Members] can raise things and speak very frankly and [...] ask us very pressing questions," Melissa said. "[This] doesn't necessarily happen with other commercial partners. The access that members get to information and to choices is completely different and we hope to hold that space."

Melissa's latest paper, written in conjunction with several READ-COOP team members, sets out the case for cooperatives as a form of AI governance. 

Sharing Transkribus' unique model

It was only in writing her new research paper that Melissa realised just how unique this model was. "It's only recently that I have come to understand that we are the only AI business that's been set up as a cooperative," she explained. "It's a very interesting model that the rest of the world should be looking at for AI and it's one of the reasons why I wanted to write the paper. I think it's important [...] to document this in a way which allows other people to understand what we've done."

Written in collaboration with academic colleagues and READ-COOP team members, The artificial intelligence cooperative: READ-COOP, Transkribus, and the benefits of shared community infrastructure for automated text recognition analyses the cooperative structure of READ-COOP and its effectiveness in sustainably supporting AI infrastructures. "There's already been a huge amount of interest in it. I've already been asked to give a keynote at the first conference about cooperatives and AI, which is going to be in Istanbul in November 2025."

You can read the paper here.

The success of Transkribus

READ-COOP has been operating for six years now, and it is clear that the cooperative model is working. In that time, the Transkribus platform has become more popular than ever, and is now used by cultural heritage institutions around the world to make historical documents accessible to everyone. For Melissa, this success is not just down to enhanced technology and usability, but also to the cooperative values that Transkribus embodies.

"It's a value system that people [who] work in cultural heritage institutions understand, which is that we're not here to make personal profit and we're not here to make our fortune. We're doing it because it's a good thing to do."

Thank you, Melissa, for taking the time to talk to us!

Want to support the world's only AI cooperative?

As a member of READ-COOP, you can help make the world's historical documents accessible to all in a way that is ethical, transparent, and governed by users instead of shareholders.

For more information about becoming a member, visit our website.