In May of this year, Taylor & Francis (which also owns Routledge) sold access to its research to Microsoft for about $10million. Taylor & Francis’ parent company confirmed to Bookseller that “it is providing Microsoft non-exclusive access to advanced learning content and data to help improve relevance and performance of AI systems”. The $10 million will cover initial access, with additional recurring payments of undisclosed sums to continue over the next three years. The announcement strikes a nerve with academics and other authors worried about the possibility of their published works being sold as training data for AI systems. As training data, these published works will serve as the raw materials that large language models draw on in order to respond to a given prompt based on statistical predictions of the most likely association.
Authors were not consulted about this deal, nor were they notified of its signing. Most learned via word of mouth after the fact. This eliminated any meaningful opportunity to opt out and exclude one’s own work from the deal. Taylor & Francis told Bookseller that they remained committed to “protecting the integrity of our authors’ work and limits on verbatim text reproduction, as well as authors’ rights to receive royalty payments in accordance with their author contracts”.
And yet, at least in IJFAB Blog Co-Editor Mercer Gary’s experience, efforts to negotiate the parameters of AI use in academic publishing contracts have thus far taken place on shaky ground. Though language protecting authorial rights and securing compensation for profits made as a result of use of the work for AI training purposes is circulating, legal teams at publishing houses have been reluctant to adopt it. Norms and policies surrounding the issue have yet to be widely established, leaving the state of play somewhat uneven and unpredictable. For instance, while Taylor & Francis has yet to assure its authors that they will receive compensation for their work’s contributions to AI training, personal communications with Oxford University Press have indicated that any future sale of research material for AI training purposes would be covered by electronic subsidiary rights royalties clauses in publishing contracts.
The Author’s Licensing and Collecting Society is currently conducting a survey about two proposed measures for protecting authorial rights. See more of ALCS Chief Executive Barbara Hayes’ remarks on coming changes with AI in publishing.
We encourage continued discussion and brainstorming in the comments about strategies for feminist bioethicists to pursue in negotiating authorial rights and protections over AI use.
