Academic publisher sells access to researchers’ articles to Microsoft AI – AU professor feels deceived

Many researchers were left shocked this July when Ruth Alison Clemens, a literary researcher from Leiden University in The Netherlands, wrote on X that she’d heard a rumour about Taylor & Francis. The rumour was that the prestigious, long-established publishing house had sold access to the publisher’s many research articles to Microsoft AI for GBP 8 million, equivalent to DKK 71 million.

And it was true. On 19 July, The Bookseller magazine reported on Taylor & Francis’ new source of income, and on 25 July they revealed that the publishers had entered into a second partnership with another AI company. The publisher’s parent company, Informa, expects the two deals to bring in DKK 513 million in 2024. According to The Bookseller, Taylor & Francis did not wish to disclose whom they had sold access to in addition to Microsoft, which is a major investor in OpenAI – the research organisation behind ChatGPT.

The sale has caused outrage among researchers, who are responsible for creating the publisher’s content in the form of articles and books – completely unpaid – yet who were not consulted about the deal. Will they be compensated for their data being used by third parties to develop AI tools? Will they be able to opt out?

One of the researchers left astonished is Peter Dalsgaard, professor of information studies at Aarhus University, who himself conducts research on artificial intelligence. Peter Dalsgaard has published a number of research articles and a book chapter in Taylor & Francis’ journals, but the news about the publisher’s new AI collaboration has made him consider whether he should drop the big publisher in the future.

“I see it as quite a major breach of trust, and I’m surprised they have the nerve to do it. We know that they have to make a profit, but this just seems an extreme course of action,” says Peter Dalsgaard.

He has received no information about the new partnership from Taylor & Francis or one of its subdivisions Routledge with whom he has published – even though, in his opinion, this represents an extremely radical choice by the publisher.

“At the time, I agreed to them publishing my research. It was on this basis I entered into collaboration with them. But the publisher has taken this one step further and said that my articles can also be used for another purpose. And that’s why I – and many of my colleagues, as I can see on social media – feel deceived, because Taylor & Francis are using the articles we submitted to them for other purposes. What’s worse is that they are selling our articles on to third parties and not even telling us who these third parties are and what they’re using our research for. And we know that there are all sorts of issues involved in large language models and the training of AI,” says Peter Dalsgaard.

Taylor & Francis: Content usage is strictly controlled 

In a written response to Omnibus, Taylor & Francis explain that collaborating with AI companies will help the publishers develop their own AI applications. However, they did not provide direct answers to a number of Omnibus' questions, including whether academic authors were notified of the collaboration, whether they will be compensated for their data being used, and whether they will have the option to withdraw their own research articles from the new collaborative agreement. They also failed to offer their stance on the fact that many researchers have expressed concern about publishing with Taylor & Francis in the future.

“Taylor & Francis is an extensive user of machine learning, AI and automated technology in areas including research submission, research integrity, plagiarism and authenticity. Like many, we are exploring new applications that will improve research and make it easier to analyse data, generate hypotheses, automate tasks, work across disciplines and research ideas. We have entered some selective partnerships with leading AI companies to accelerate this work, which includes providing access to a range of archive content for research and training purposes, supporting the accuracy and relevance ofAI models, particularly LLMs (large language models, ed.). As with all our licensing activities, content usage is strictly controlled and rights holders will receive royalties in line with their contracts,” Taylor & Francis state in their written response.

Peter Dalsgaard: We’re losing one of the cornerstones of academia

Peter Dalsgaard is also open to the possibilities that AI technologies present to research, he explains.

“It’s one of the areas where I see great potential. Being able to recognise patterns in vast amounts of data and to summarise and analyse across this data. We already have tools that can really help us in our research. But when I use these tools, as a researcher, I always ask myself how much I can trust their output and how they work. Are there things that are being hidden from me? Is the information I’m being shown representative? For researchers, these tools give rise to important methodological considerations,” he says.

The professor also emphasises that it’s researchers who create the content but the publishers who cream off the profits.

“Let’s just say that the publishers give the tech companies the opportunity to help them develop AI technologies and develop new services for the publishers, which we’ll also have to pay to use. So we supply the content and the editorial work they use to train the AI systems that they’ll sell back to us. It’s a vicious circle,” he says.

With the sale of their articles to a third party, researchers lose the control and transparency they’ve had until now, argues Peter Dalsgaard.

“Today, we have traditional ways of knowing how our articles are being used. If somebody refers to my work, there are clear rules for how they must cite me and my article. One of the cornerstones of the academic world is that you can trace things back to their source. This is notoriously difficult to do using large language models. We’re losing one of the cornerstones of academia now that the publishers are giving other people access to our content in ways that we cannot inspect or object to,” he says.

Collective mobilisation

According to Peter Dalsgaard, this situation makes it clear that researchers need to think “more seriously” about the rights they give publishers over their work. The Taylor & Francis example clearly shows that publishers are well aware of the power they have, he believes.

“We have to find out whether we have the option of withdrawing our consent at some point. Whether we need to include clauses that say the publishers can only use our research for certain purposes and that, if they want to use it for other reasons, we need to give our consent. Looking back now, it seems as though we just signed something and inadvertently gave the publishers the right to use our work,” he says.

“We took it for granted that we were giving them the right to publish our article in a journal of our own choosing, for which our colleagues had provided editorial work in good faith. Now it transpires that they can use our articles for all sorts of purposes. And they may well have the law on their side, but in my opinion it’s very ethically problematic,” says Peter Dalsgaard.

Peter Dalsgaard is a member of the academic council at Faculty of Arts, and he intends to raise the issue at a council meeting. He is also considering how researchers can join forces to tackle this problem on a more general level. But he is aware that he is speaking from a privileged position in the research world, which not everybody has. For early career researchers, publishing in prestigious journals is one of the ways of getting to the top.

“It’s easier for me to say as a senior researcher. But I also think that it’s people like me who need to do something about it. Because, if you’re an early career researcher, your career is partly determined by the publication structure and the system of prestige built into it. It’s difficult for me to say to a PhD student or a postdoc that they should stop publishing in this or that prestigious journal, because their career depends on it,” he says and continues:

“This situation requires some form of collective mobilisation or potential boycott. We occasionally see, for example, that an entire editorial team resigns from a journal because it disagrees with the publishers on a fundamental issue. This type of action really makes a difference in academic circles,” says Peter Dalsgaard, who emphasises that such protests can only work if the researchers have alternative publishing channels ready.

Wiley also profits from AI partnerships

There are signs that publishing houses collaborating with tech companies might be becoming a trend. On 29 July, the Inside Higher Ed news website reported that Wiley had also entered into partnerships with AI firms. In a press release from June, Wiley stated that it had completed a project with a “large tech company” and that a second project would be realised in 2025.

In response to Taylor & Francis’ sale of articles to AI companies, the largest trade union for professional authors in the UK issued a press release in which it criticised the deal.

“We are very concerned to see publishers signing deals with tech companies without consulting their authors and creators first. There are substantial copyright, moral rights and data protection questions that need to be addressed, as well as ethical questions about transparency and fairness of payment from (authorised) uses of creators’ works by AI tech companies,” the union writes.

Several researchers have confirmed on X that they haven’t heard anything from Taylor & Francis. Professor of anthropology at the University of Oxford Eben Kirksey, for example, writes:

“Um, pay us please? Taylor and Francis is making millions by selling academic articles to Microsoft AI. At least give scholars and published authors the right of opting out of this money making data grab.”

Olav W. Bertelsen, associate professor and joint union representative for academic staff at AU’s Faculty of Natural Sciences, writes on X that it is “time for academics, universities, and states worldwide to take full control over academic publishing. The commercial publishers have become unnecessary parasites.”

According to The Bookseller, Taylor & Francis achieved a turnover of DKK 2.6 billion in the first six months of this financial year. Its parent company, Informa, expects to have a turnover of over DKK 8 billion in 2024.