Industries / Media, arts and recreation / Media and entertainment / Publishing
Rights
Rights work decides who may copy, translate, adapt or license a text, where and for how long, and makes sure the author and publisher are paid when someone does.
6
Milestones achieved, from 1886 to 2024.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1886
Copyright crosses borders
The Berne Convention, the first multilateral copyright treaty, was signed in Berne, Switzerland, on 9 September 1886. It requires each member country to protect works from other members as it protects its own, with no registration or other formality, and its current text sets a minimum term of the author's life plus 50 years. The United States attended the conference but did not sign.
Credit
the founding member states
1891
America protects foreign authors
The International Copyright Act, signed by President Benjamin Harrison on 4 March 1891, gave the works of foreign authors copyright in the United States for the first time. The first foreign work registered was Saints and Sinners, by the English dramatist Henry Arthur Jones, on 3 July 1891.
LessonProtection for foreign authors came only after more than half a century of public debate and lobbying.
Credit
US Congress; President Benjamin Harrison
1982
Frankfurt turns to rights
The Frankfurt Book Fair in its modern form began in September 1949 with 205 German publishers in the city's St Paul's Church. By 1982, 104 literary agents registered for it, which the fair's own history reads as a sign that it was "increasingly turning into a venue for licensing deals"; by 1983 its Literary Agents Centre had become "a marketplace for planned bestsellers and international marketing".
Credit
Frankfurter Buchmesse
2019
Authors get reports and exits
The EU's copyright directive for the digital single market, Directive (EU) 2019/790, requires anyone who licenses or buys an author's rights to report at least once a year how the work is used and what it earns (Article 19), and lets the author take back exclusive rights in a work that is not being used (Article 22). Its Article 4 also allows text and data mining, the automated copying of works to analyse them, unless the rights holder has reserved those rights, by machine-readable means for works online.
Credit
European Parliament and Council of the European Union
2022
A shadow library seized
US prosecutors charged two Russian nationals, Anton Napolsky and Valeriia Ermakova, with criminal copyright infringement, wire fraud and money laundering for running Z-Library, a site that claimed to offer more than 11 million e-books, and seized its network of about 249 web domains. The pair were arrested in Córdoba, Argentina, on 3 November 2022. Days later, readers reported that the site could still be reached on the dark web.
LessonSeizing domains took the site off the open web but did not remove the files, which stayed reachable by other routes.
Credit
US Attorney's Office for the Eastern District of New York; FBI
2024
Paid licences for AI training
HarperCollins asked thousands of its authors in November 2024 to opt nonfiction books into a licence for training an AI model, for $5,000 a title split equally between author and publisher, with output limited to 200 consecutive words or 5% of a book's text and a pledge by the AI company not to scrape piracy sites. The Authors Guild welcomed the principle that authors must agree, but said a 50-50 split gave the publisher too much.
Credit
HarperCollins; the Authors Guild
Watch for
AI translation for self-publishers
Amazon launched Kindle Translate in November 2025, a free AI translation service, in beta, for authors on Kindle Direct Publishing; by June 2026 it translated between English and Spanish, German, French, Italian and Portuguese. Amazon says every translation is automatically checked for accuracy and quality before publication, that thousands of authors have used it and that readers rate the books 4 stars or more on average. It also says fewer than 5% of titles on Amazon.com have historically been available in more than one language.
Achieved whensomeone outside Amazon compares a sample of Kindle Translate books with human translations of the same texts, judged blind by readers of the target language.
Claimed by
Amazon, November 2025 and June 2026
Machine-first romance translation
HarperCollins France confirmed in January 2026 that it is "conducting tests" with Fluent Planet, a Bordeaux agency whose translators use AI tools for part of their work, on its Harlequin romance series, citing falling sales and a €4.99 price for the Azur series; it said no Harlequin series had been translated solely by machine. The French literary translators' association, ATLF, said the method runs texts through automatic translation and pays freelancers to correct the output. 404 Media reported that Fluent Planet claims its AI gives translations of "2 to 3 times higher quality" than other large language models.
Achieved whensomeone other than the two companies publishes blind ratings that compare these translations with human ones.
Claimed by
HarperCollins France and Fluent Planet, December 2025 to January 2026
Open
01
Recount the missing decades
A fresh count of which old books can still be bought new. Draw a random sample of books, find the decade each was first published, and record whether it is on sale new in print, through print-on-demand (copies printed one at a time when ordered) or as an e-book, so that books still in copyright can be compared with books in the public domain.
The briefCredit · unclaimed02
Restart the translation index
A yearly count of books translated into and out of each language, country by country, from 2010 onward, built from the records national libraries already keep, where a record can carry a code for the language a book was translated from.
The briefCredit · unclaimed
Open 01 · the brief
Recount the missing decades
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A fresh count of which old books can still be bought new. Draw a random sample of books, find the decade each was first published, and record whether it is on sale new in print, through print-on-demand (copies printed one at a time when ordered) or as an e-book, so that books still in copyright can be compared with books in the public domain.
Why it is still open
Paul Heald's study, published in 2014, drew a random sample of 2,266 new fiction titles on sale at Amazon and found more books on sale from the 1880s than from the 1980s; his look at e-book markets, used books and a public library suggested that no other marketplace for out-of-print books had yet emerged. His later studies measured how books come back into print when rights revert to their authors, using samples of bestsellers and reviewed books (2017), and reported data from the UK and Canadian book markets (2020). Print-on-demand has grown since the 2014 sample: Ingram's print-on-demand library alone held more than 18 million titles by 2020, by Publishers Weekly's account. This research found no rerun of the random sample on today's market.
gwern.netcopyrightevidence.orgcopyrightevidence.orgcopyrightevidence.orgpublishersweekly.com
Built with what exists
Open Library and Wikidata for each title's first publication date; national library catalogues; listings at Bookshop.org, ebook shops and publishers' own sites, recorded by hand or by a web agent working within each site's terms (see the Note "Deleting 80% to start over"); Python.
First step
Draw 200 titles by Heald's method and record by hand whether each is on sale new.
Done looks like
A random sample of at least 2,000 titles, counted by decade of first publication, separately for new print, print-on-demand and e-book editions, plotted against Heald's curve, with data and code published, plus a second sample from a store in a country with a different copyright term.
Open 02 · the brief
Restart the translation index
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A yearly count of books translated into and out of each language, country by country, from 2010 onward, built from the records national libraries already keep, where a record can carry a code for the language a book was translated from.
Why it is still open
UNESCO's Index Translationum, the world bibliography of translations, stopped being kept up around 2012. By an analysis that Péter Király and András Kiséry published in December 2024, only two countries' data for 2011 were processed, 2010 was the last processed year for nine, and 2009 for 29. The databases that remain each cover one place or one language, such as the "3%" database of literary translations published in the US, now hosted by Publishers Weekly. Eurolit, a project at the University of Siegen funded by the German Research Foundation, gathers translation flows within Europe from national deposit libraries, but its data was not public when Király and Kiséry wrote, and their own database covers translations from Hungarian from 1800 to 2009. They write that no one has global figures now.
Built with what exists
Open bibliographic data from national libraries such as the Deutsche Nationalbibliothek and the British Library; the language field of MARC 21 library records (field 041), whose subfield h holds the original language; an AI model to classify records that lack the code, checked by hand on samples; Python.
First step
Download one year of one national library's records and count the translations by original language.
Done looks like
Counts by source language and year for at least ten countries, each with 200 records checked by hand and at least 95% classified correctly; data, code and checks published openly.