The latest installment of Clarke & Esposito’s The Brief newsletter is, as always, sharp and informative. The lead bit is on Google’s new Gemini, and its example AI academic search. Thanks to Google Scholar, Gemini can mine OA content at the article level, even in hybrid journals—and also scrape green AAM versions:

The implications of this granular indexing on licensing arrangements is worth noting. As more and more articles are published on an OA basis, even if in hybrid journals or via green OA, Google Gemini (and other generative AI platforms) will have more free-to-them content to incorporate into their platforms. This raises the question as to whether liberal reuse licenses might need to be revisited by publishers.

This point—publishers, roll back your liberal OA licensing!—was made by Joe Esposito back in July at the Scholarly Kitchen (“What publishers need is more copyright protection, not less”).

In the words of the latest Brief:

The CC BY license is looking increasingly like a mechanism to transfer value from scientific and scholarly publishers to the world’s wealthiest tech companies.

Maybe. But here’s a rewrite: The scholarly publishing business is looking increasingly like a mechanism to transfer value from scholars, taxpayers, and universities to the world’s most profitable companies.

Sam Moore, concluding a thoughtful analysis of rights-retention policies in Europe an elsewhere:

To be clear, very little of the negotiations around rights retention represents much more than liberal market economics—it should not be confused for the kind of political radicalism that higher education institutions are unable to perform. However, this is not its point: rights retention makes institutions more combative towards extractive practices in the publishing industry by demanding certain transactional conditions that are about more than just price. To that extent, rights retention can be well coupled with a strategy that supports and nurtures alternative and more radical approaches to open access that are less dependent on market incentives.

Roy Kaufman, CEO of the for-profit Copyright Clearance Center (CCC), posted excerpts from the company’s response to the US Copyright Office’s call for comments on AI training data. The Scholarly Kitchen post is fascinating and worthy of notice. It’s hardly surprising that the CCC—the publishing industry’s main licensing and permissions broker—took a muscular stance:

There is certainly enough copyrightable material available under license to build reliable, workable, and trustworthy AI. Just because a developer wants to use “everything” does not mean it needs to do so, is entitled to do so, or has the right to do so. Nor should governments and courts twist or modify the law to accommodate them.

What’s more interesting is CCC‘s frontal assault on the fair use rationale that big tech has used to justify its permission-free training-data hoovering. The key passages are a bit technical, filled with references to “transformer architecture” and “tokenizers.” The plain reason is to establish that language-model training requires large swaths of text:

It is important to understand the role of word embeddings when it comes to copyright because the embeddings are the representations (or encodings) of entire sentences, paragraphs, and even documents, in a high-dimensional vector space. It is through the embeddings that the AI system captures and stores the meaning and the relationships of the words from the natural language.

Such embeddings, CCC continues, are used in “practically every task” in generative AI. The message? OpenAI and the rest aren’t using mere snippets.

There’s lots more in the comments, which are (I predict) a preview of how the big commercial publishers will approach litigation and/or licensing negotiations with the Silicon Valley model builders.

Can the likes of OpenAI scrape up copyrighted content into their models, without permission or compensation? The tech companies think so; they’re fresh converts to fair-use maximalism, as revealed by their public comments filed with the US Copyright Office. The companies’ “overall message,“ reported The Verge in a round-up, is that they “don’t think they should have to pay to train AI models on copyrighted work.”

It’s time to heat up the popcorn.

Speaking of APC revenue, Leigh-Ann Butler and colleagues have a great paper out in Quantitative Science Studies, on APC revenue from the big five commercial publishers (2015–2018):

Revenue from gold OA amounted to $612.5 million, while $448.3 million was obtained for publishing OA in hybrid journals. Among the five publishers, Springer-Nature made the most revenue from OA ($589.7 million), followed by Elsevier ($221.4 million), Wiley ($114.3 million), Taylor & Francis ($76.8 million) and Sage ($31.6 million).

That’s over $1 billion dollars, with nearly $200 million flowing to just two SpringerNature mega-journals (Scientific Reports and Nature Communications).

Science has a solid news story on the study. The Science piece includes self-serving complaints from SpringerNature on the study’s methods. One of the author’s, Stefanie Haustein, is quoted in reply:

If publishers believe our estimates are not accurate, we would appreciate it if they would publish their data and be transparent. Their lack of transparency is precisely what has made our work so slow and difficult.

As John Milton put in the 17th century: “They who have put out the people’s eyes, reproach them of their blindness.”

The just-released U.S. Office of Science and Technology Policy (OSTP) report on OA financing is definitely interesting—it’s far more in-depth in its scope than the last year’s Nelson Memo. References to cOAlition S, diamond OA, subscribe-to-open, and other funding-landscape arcana, for example, litter the new report. And the report’s authors, among many other things, tried to estimate how much US taxpayers are forking over for APCs. Their guess is $375 million a year in 2021, a $100 million increase over five years.

Another fascinating (if unsurprising) detail on publisher concentration:

In the U.S., five publishers — Elsevier, Springer Nature, Wiley, the American Chemical Society (ACS), and Oxford University Press (OUP) — account for 51.4 percent of the volume of federally funded publications from 2016 to 2021, according to data retrieved from Clarivate’s Web of Science. Similarly, five publishers — Elsevier, Spring [sic] Nature, Wiley, OUP, and MDPI — account for 51.4 percent of the volume of openly accessible federally funded articles from 2016 to 2021.

The report, for all its taxonomic detail, is a disappointment. That’s because the bottom-line conclusions haven’t changed: the report endorses the same non-committal APC-friendly stance as the Nelson Memo. This despite earnest paragraphs on “inequities in publishing” and due mention of the author-excluding implication of APCs.

A coda to Tuesday’s post on SPARC’s Elsevier report. Here’s Sam Biddle, writing for The Intercept on Elsevier corporate sibling LexisNexis Risk Solutions:

The popular data broker LexisNexis began selling face recognition services and personal location data to U.S. Customs and Border Protection late last year, according to contract documents obtained through a Freedom of Information Act request.

According to the documents, obtained by the advocacy group Just Futures Law and shared with The Intercept, LexisNexis Risk Solutions began selling surveillance tools to the border enforcement agency in December 2022. The $15.9 million contract includes a broad menu of powerful tools for locating individuals throughout the United States using a vast array of personal data, much of it obtained and used without judicial oversight.

Are any of the LexisNexis tools on offer drawing on data gathered through Elsevier’s ScienceDirect? Who knows—there’s no real answer in the contract documents nor the ass-covering LexisNexis FAQ page.