Such hyperlocal histories are a crucial resource, a way for particular communities to preserve and chronicle their cultures, as well as a means for marketing their regions to tourists and chance visitors. But their audiences are consequentially limited, so Arcadia does not usually approach its authors with hundreds of dollars on offer. In its email to Brown, the publishing house even pointed out that these opportunities for author compensation “could be very limited in the future,” pointing to summertime court verdicts that recognized the A.I. training process as fair use—even with copyright material. Arcadia was offering its authors a favor, while making clear it didn’t have to, and pointing out that this could be their only chance…
Arcadia is hardly the only book publisher to ink such opaque contracts with the A.I. overlords, despite the spirited objections and lawsuits brought by various authors. University and scholarly presses—which have been confronting the fallout from the Trump administration’s mass grant cancellations, higher printing and shipping costs from tariffs, and industry headwinds—are providing the model. Taylor & Francis, an academic publisher based in the United Kingdom, signed a $10 million deal with Microsoft last year to share a portion of its catalog for A.I. training, in exchange for annual payments from the tech giant through 2027. Authors were reportedly given no notice and their royalties were measly in turn; Bloomberg quoted one anonymous Taylor & Francis author who claimed to earn only $97 for ceding their book to the training maw. (A T&F spokesman told Bloomberg that the payments were “in accordance with the licensing terms and royalty statement periods in their contracts,” while parent company Informa declared in a press release that “the agreement protects intellectual property rights.”) Wiley, an over-200-year-old academic publishing house, has already struck multiple A.I. deals for licensing and product integration, offering up its works to inform the output of Perplexity’s LLM and Amazon Web Services’ chatbot.
For the publishers, the arrangements were lucrative. For the authors, the payouts were much less so. In July, Johns Hopkins University Press gave the authors of its 3,000-title catalog an Aug. 31 deadline to opt out of having their works become A.I. training fodder in a new tech partnership. If they opted in, they would receive a little under $100 per work. Like Arcadia, Hopkins Press did not disclose the A.I. company involved or the money it was hoping to earn. It did press the urgency of signing now while writers still had some agency, and reminded them who here really has the power. “In your contract, you provide us with the rights to go ahead with this kind of licensing,” Barbara Kline Pope, executive director of Hopkins Press, wrote to her writers. “However, we would like you to have the ability to opt out if you so choose.” The press was not suffering businesswise, she clarified, but it was “exploring how our financial model may need to evolve.” One author who went for the opt-out contract addendum with Johns Hopkins Press shared the resultant language with Inside Higher Ed; it warned that “sales and reach” of their work might suffer due to the A.I. opt-out…
A lot is still unclear, but a few things are apparent: A.I. companies are aggressively reaching out to book publishers to strike deals that will allow them to sidestep the litigation that led to the Anthropic settlement and avoid the heftier payouts. Whichever unnamed firm approached Arcadia, it took a particular interest in the wordier History Press, indicating that generative text remains the lodestar. And if the Theodore/Franklin Roosevelt mix-up is representative of other chatbot hallucinations, that perhaps indicates the need not just for these bots to brush up on history and text, but to ramp up the representation of local history in the mix in order to make the LLMs more universal.
It sounds like AI companies want large bodies of texts and academic publishing provides that.
It might just be about the words and texts but I wonder if any of the AI services actually wants the research information. Imagine one of them builds and advertises a specialty in local history. To look for local history online right now might require some digging (see steps for investigating suburbs here and here). What sources to trust? Where can I find specific information about people and places?
For example, I was recently looking at the different presentations about suburban communities between Wikipedia and Grokipedia. In some ways, the pages were similar in terms of their headings and the kinds of information presented. However, they drew on some different sources. Does a community’s website provide the best overview of a community? Where might published histories fit? Who can incorporate “official” overviews and the lived experiences of residents and those studying the history?
Perhaps there would be a market for accurate local history AI. Would it help people doing genealogies or interested in local development or looking to move to a new place?
Legally Sociable 