You have /5 articles left.
Sign up for a free account or log in.
![Books in a digital space](/https/www.insidehighered.com/sites/default/files/styles/max_650x650/public/2024-07/GettyImages-646946012.jpg?itok=-sGe4Buv)
Legal experts say giving AI access to scholarly research is likely protected under publishers’ current licensing agreements with authors.
SvetaZi/iStock/Getty Images
Academic researchers around the world are reeling from news announced in May that Informa, the parent company of academic publisher Taylor & Francis, has signed a $10 million data-access agreement with Microsoft.
The AI partnership agreement gives Microsoft “nonexclusive access to Advanced Learning Content” across Taylor & Francis’s nearly 3,000 academic journals. After the initial access fee of $10 million, Informa said it would receive recurring payments for the next three years.
Scholars across the U.S. and Europe said the news took them by surprise and they’re worried about how their research will be republished and cited by the publisher’s AI tools.
“I was shocked, not that it had happened, but that no one had said anything to us at all,” said Lauren Barbeau, assistant director of learning and technology initiatives at Georgia Institute of Technology, who has also published a book with Routledge, an academic publisher owned by Taylor & Francis.
“I’ve come to terms with the fact, as an author who has published, that at some point my work is going to go into AI, whether that’s through an illegal copy published somewhere on the internet or some other means. I just didn’t expect it to be my publisher.”
Barbeau said what disturbs her the most “is that there’s so little agency given to the authors about how their work is going to be shared and presented to the world,” which underscores the already “predatory” nature of academic publishing.
She admits she’s optimistic that the deal means well-researched, accurate data will be used to train AI models. “Then, there’s the other side of me that wonders, if it’s going into AI, is it going to properly contextualize my work?” she said. “Or is it going to take my words and my thoughts out of context and misrepresent them because it’s not giving the full picture?”
Ruth Alison Clemens, a lecturer in literature at the Leiden University Centre for the Arts in Society in the Netherlands, wrote on X, “The bulk sale of data to big tech to train AI that has as-yet unknown reach & applications sets a new and worrying precedent for data capitalism & academic research. So email your editors!”
The deal could also mean the potential loss of citations, which are valuable currency in a scholar’s career.
“What we’ve seen of AI by and large is that it’s not very good at citing or articulating what came from where and why,” Lance Eaton, a higher education consultant who specializes in generative AI, said. “Here, the publisher is selling the content to an AI tool, which we don’t know how it’s going to be used, but we know they want the scholarly information. What happens to those ideas in that scholarship? How does it get used and how do people get credit for it?”
A news release from Informa said it’s considering such concerns, and the agreement “protects intellectual property rights, including limits on verbatim text extracts and alignment on the importance of detailed citation references.”
$75 Million-Plus Profit Expected
Informa’s half-year results published on July 24 said the firm had “secured a second major partnership” with another AI company and expects AI-related revenues to exceed $75 million for the year.
The data-access agreements, it said, “are a source of significant new value for Taylor & Francis and additional royalties for authors.”
Academics, however, are skeptical that they will financially benefit from the agreement. While Barbeau—like most scholars—earns single-digit royalties when publishers sell copies of her book, she said, “It’s still a concern that my work is now out there and publicly accessible and they were reimbursed for that.”
She wants to see changes in the way research is evaluated and shared. “This is not the beginning of this conversation, but I hope it’s a tipping point.”
Meanwhile, Eben Kirksey a professor of anthropology at the University of Oxford, wrote on X, “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.”
Giving scholars the opportunity to opt in—or out—of their data being used to train Microsoft AI is critical to defending their agency, said Heather Joseph, executive director of SPARC, the Scholarly Publishing and Academic Resources Coalition, a nonprofit group whose members include libraries and academic organizations advocating for open-access research.
“It’s an unqualified good thing for generative AI to be trained on the latest and most high-quality information,” she said. “But there are some circumstances under which we don’t want our information to go to a particular use.”
Developing standards on how and when authors can both retain control of their content and allow other entities to benefit from it is what Joseph is focused on in the wake of the Taylor & Francis deal: “It’s a balancing act at this point.”
AI to ‘Improve Research’
Taylor & Francis publishes over 2,700 journals and partners with more than 700 scholarly societies, according to its website. It owns Routledge, a leading publisher in humanities and social science fields, as well as CRC Press, a publisher of medical, science and technology imprints.
Mark Robinson, a spokesperson for the publisher, didn’t answer specific questions about why the company didn’t contact its authors prior to the announcement of the AI partnerships.
However, he said in an email that the deal is intended to accelerate the company’s ongoing exploration of “new applications that will improve research and make it easier to analyze data, generate hypotheses, automate tasks, work across disciplines and research ideas.”
Informa has already been using AI to maximize efficiencies in its academic markets, including with research submissions, authenticity screening, validation and accuracy, customer response, and content summaries.
The contract with Microsoft, which is set to last until 2027, “expands on our existing enterprise technology partnership to explore how the use of AI can drive innovation and improve productivity across the Informa Group,” according to a news release from the company.
The partnership is focused on improving research productivity, developing—and improving the speed and accuracy of—automated citation referencing, testing the development of expert AI agents that can assist authors and librarians with research and the creation and sharing of new knowledge, and improving relevance and performance of AI systems.
Legal Recourse Not Likely
The scholarly publishing industry, even before the proliferation of AI, often disproportionately favors publishers over individual authors.
For many academics, “publishing is the key to their tenure,” said Eaton, the generative AI–focused higher ed consultant who’s also writing a dissertation related to the ethics of scholarly publishing. “It’s really hard to think about what leverage they have when they’re often publishing for free.”
Additionally, top-tier academic research is often only available through costly subscriptions—some worth hundreds of thousands of dollars—that not all academic libraries can afford to buy, pushing scholars to turn to academic piracy networks, such as Sci-hub or Libgen, to access relevant research papers.
Although the Taylor & Francis deal with Microsoft may be among the first of its kind – Wiley announced a similar deal with an unnamed "large tech company" last month – Eaton said he expects more scholarly publishers to sign similar data-access deals, which will only worsen the power imbalances in academic publishing.
“Those scholars feel disempowered by the way that the scholarly publishing industry strips them of their ability to do the work they are paying to do through their education, and they have to find ways to pay or break the law in order to access research to continue their work,” he said.
The AI piece is just one more insult, he said.
“Now, when scholars do publish, not only do they have to give all their rights away to these publishers, the [publishers] are going to profit even more.”
While authors say they understand that licensing agreements allow publishers to sell access to their work, some have questioned if those policies also include AI systems.
But legal experts believe that, although it’s dependent on the language of a particular copyright agreement, most publishers who sell data to generative AI tools will likely be insulated from legal consequences.
“The typical copyright agreement major publishers use is almost certainly going to make it impossible for academic authors to sue,” said James Grimmelmann, a professor of digital and information law at Cornell Law School. “The AI training is going to be fairly similar to the ways in which the publishers already sell bulk access to their collections to libraries and other institutions.”