top of page
RESOURCES FOR WRITERS

Literary Translators Challenge AI in a Fight for the Future of Authorship

6 hours ago
12 min read
As publishers experiment with automated translation, literary translators have joined a major copyright battle against OpenAI and Microsoft. Their intervention exposes a profession already struggling with low pay, uncertain ownership rights, and the growing expectation that literary judgment can be replaced by software.

The American Literary Translators Association has entered a federal copyright dispute against OpenAI and Microsoft at a moment when the publishing industry is beginning to reconsider how much of a translator's work requires a translator. The organization's intervention, part of a legal filing submitted October 30 by a coalition of 12 professional associations, comes amid expanding experiments with artificial intelligence in commercial publishing. Some publishers are testing systems that generate translations for human professionals to revise. Translators fear that the arrangement could leave them responsible for the quality of a finished book while reducing their authority over its language, their compensation, and their claim to authorship.

The filing, which received renewed attention in Publishers Weekly on November 18, supports authors pursuing copyright claims over the use of their books to train generative AI systems. The plaintiffs argue that OpenAI and Microsoft have exploited copyrighted material without authorization, threatening the market for work produced by professional writers. The defendants maintain that their training practices are protected by fair use. For literary translators, the dispute reaches into a profession whose creative contributions have long been less visible than those of the authors they translate, and whose economic position was precarious well before artificial intelligence entered the industry.¹

The disagreement arrives at a consequential point in publishing's adoption of automated tools. HarperCollins has discussed plans to increase automation in overseas translation operations; its French division has tested AI-assisted translation for its Harlequin imprint. Specialized companies are offering publishers machine-generated first drafts followed by human revision, presenting the process as a way to reduce costs and bring more books into additional languages. Translators have responded with concerns about professional displacement, the quality of the finished work, and the possibility that the person responsible for transforming a novel's language could be treated as an editor of software output rather than an author in their own right.²

The Copyright Dispute

The American Literary Translators Association, founded in 1978, represents practitioners working across fiction, poetry, drama, and literary nonfiction. Its participation in the consolidated litigation, In re OpenAI, Inc., Copyright Infringement Litigation, places translators alongside organizations representing journalists, novelists, photographers, playwrights, and other creative professionals. The coalition's amicus curiae brief supports the author plaintiffs' motion for summary judgment, asking the court to consider how unlicensed use of copyrighted works for AI training could affect the markets that sustain professional creative labor. ALTA and its fellow signatories are presenting arguments to the court rather than pursuing individual damages through the filing.¹

The underlying cases include litigation brought by the Authors Guild and individual authors beginning in 2023. The consolidated proceedings, before Judge Sidney H. Stein in the Southern District of New York, concern allegations that copyrighted books and other material were copied to train AI models without authorization. The plaintiffs contend that this conduct infringes their rights and that the resulting technology threatens established and potential markets for their work. OpenAI and Microsoft dispute the infringement claims and argue that relevant uses of copyrighted material for model training qualify as fair use. Both sides have sought summary judgment on substantial issues, asking the court to resolve legal questions without waiting for a full trial.³

ALTA's argument concerns the particular position of translators within that dispute. An authorized literary translation can itself be protected by copyright because the translator contributes original expression to a work whose underlying rights belong to someone else. The selection of words, arrangement of sentences, handling of cultural references, and recreation of an author's voice involve creative decisions that may differ substantially from one translator to another. ALTA argues that automated systems developed through extensive use of copyrighted material could undermine the market for those contributions, particularly if publishers come to regard human translation primarily as the correction of machine-generated drafts.

The concern extends beyond the ownership of source material. A publisher commissioning a translation ordinarily engages a professional to produce a new literary text under an agreement governing payment, credit, and rights. If software generates the initial version, the publisher may instead commission a human to check and revise it, potentially changing the contractual understanding of the work. A translator could remain responsible for resolving the most difficult passages while receiving a lower fee and less recognition for the finished book. The legal status of such work would depend on its human creative contributions and the applicable contract, questions that existing publishing practices have not settled uniformly.

The coalition's position is supported by other organizations that depend upon copyright licensing. In a separate filing, the Association of American Publishers and allied groups argued that allowing technology companies to use protected works without permission could undermine both the market for original books and publishers' ability to license their content. These arguments reflect a larger dispute over whether the commercial development of generative AI should require negotiated access to copyrighted works, and whether existing copyright principles adequately address the economic consequences of systems capable of producing new text at enormous scale.⁴

Translation as Authorship

Katrine Øgaard Jensen, ALTA's executive director, has compared literary translation to a musician's interpretation of an existing composition. In an interview with Publishers Weekly, she invoked Jimi Hendrix's recording of Bob Dylan's All Along the Watchtower, asking readers to imagine Hendrix being handed an AI-generated arrangement to correct rather than developing the interpretation for which his recording became celebrated. The analogy captures a distinction translators have spent decades trying to establish in contracts and public recognition. The original work supplies the composition, but the interpreter's decisions determine the form in which a new audience experiences it.²

The differences between literary translations can be substantial. Emily Wilson's 2017 translation of Homer's Odyssey, for example, drew attention for its direct language, disciplined meter, and willingness to reconsider expressions that earlier English translators had rendered through familiar heroic conventions. Her decisions affected characterization and tone as well as the reader's understanding of the ancient poem. Such work requires sustained interpretation of a text whose meanings have been debated for centuries. Different translators produce different Odysseys because their readings of Homer, and their judgments about English prose and verse, are not identical.

The distinction becomes especially important when a literary work resists the conventions of its original language. A novel may rely upon regional idioms, unstable narration, deliberate repetition, or a sentence whose syntax creates an effect that cannot be reproduced literally. A translator must decide how those features operate and which aspects of the original can survive their movement into another language. Correcting grammar is frequently the least difficult part of the undertaking. An apparently fluent translation may communicate the events of a story while erasing the linguistic qualities that make the book distinctive.

Elisabeth Jaquette, an Arabic literary translator and executive director of the literary magazine Words Without Borders, discussed this problem in an October interview at the Frankfurt Book Fair. AI-generated English, she observed, can appear convincing even when it misunderstands an ordinary phrase in Arabic. Its fluency may conceal the error rather than expose it. Jaquette also described her experience reviewing machine translations as slower and more frustrating than translating directly from the original, because revising an existing draft can require dismantling the choices already imposed upon the text.⁵

Jaquette's experience does not establish that every machine-assisted translation takes longer or produces inferior results. Performance varies by language, genre, model, and the extent of human revision. It does, however, challenge the assumption that generating a complete first draft necessarily reduces the labor involved in producing a publishable book. The apparent speed of an automated system measures the production of text, while the translator's task includes determining whether that text preserves the meaning, style, and intentions of its source. A fluent sentence can be mistaken in ways that are difficult for a reader unfamiliar with the original language to detect.

The problem also affects how editors assess the finished work. A publisher can evaluate an English-language manuscript for coherence, pacing, and readability without necessarily knowing whether those qualities reflect the original. A translation may satisfy ordinary editorial expectations while obscuring the source author's stylistic peculiarities. Professional translators occupy the space between those competing requirements, accountable both to the writing from which they work and to the readers for whom they produce a new text. Removing their authority from the initial stages of the process would alter a relationship that has shaped the international circulation of literature for generations.

The Economics of Translation

The financial pressure to automate is readily apparent in publishing decisions already made. In January 2026, HarperCollins France confirmed that it was testing AI-assisted translation through Fluent Planet for its Harlequin imprint. The announcement followed complaints from the French Literary Translators Association and the translators' collective En Chair et en Os, which alleged that contracts were being terminated as work moved toward a model involving automated translation and human revision. HarperCollins France said that experienced translators were participating in the tests and that no Harlequin collection had been translated solely by machine-generated text.⁶

The publisher cited declining sales in the French market and the difficulty of maintaining inexpensive romance editions. Its statement referred to books in the Azur series selling for €4.99, a price point that leaves limited room for additional production costs. The experiment illustrates the financial reasoning behind automated translation more clearly than predictions about the technology's eventual capabilities. Publishers working with low-priced editions and narrow margins may view reduced translation expenditure as a means of keeping a series commercially viable. Translators, particularly those who have already lost contracts, see an immediate threat to work that previously required human specialists.⁶

Other companies have developed services around the same premise. Sweden's Nuanxed produces initial translations using proprietary AI systems before employing human translators, editors, and proofreaders to prepare books for publication. Its business model addresses a longstanding obstacle in international publishing: many books never reach foreign-language audiences because the expected sales do not justify the cost of translation. Reducing that expense could allow publishers to acquire titles that would otherwise remain unavailable. The question for literary translators is how those savings will be distributed and whether new commissions will compensate for work displaced by automation.²

The profession's existing financial condition makes those questions particularly urgent. In March, PEN America published Fairness in Publishing: The State of Literary Translation in the U.S., based on interviews with 17 translators and 20 publishing editors, together with earlier industry research. The report documented persistent problems involving compensation, attribution, royalties, and contractual rights. Citing a 2022 Authors Guild survey, it noted an average payment of approximately 13 cents per word, with the largest group of respondents reporting rates of 10 cents or less. Those figures vary considerably among languages, publishers, projects, and individual agreements, but they demonstrate the modest earnings available through work requiring substantial expertise.⁷

At 13 cents per word, a 90,000-word novel would produce a fee of $11,700 before taxes and expenses. The project might require months of drafting, research, revision, and consultation. Some translators negotiate advances against royalties, while others work for flat fees that offer no additional compensation if a book becomes successful. PEN America's research found that translators continued to encounter inconsistent crediting practices and limited bargaining power, despite the importance of their work to international publishing. A reduction in rates for AI-assisted editing would enter a profession in which many practitioners already struggle to support themselves through literary work alone.

The situation is particularly difficult for independent publishers, which account for a substantial share of translated literature in the United States. Their commitment to international writing often exceeds their financial resources. Acquiring rights, commissioning a translation, editing the manuscript, and marketing an unfamiliar author can represent considerable expense relative to expected sales. Grants and support from cultural organizations sometimes make projects possible, but such assistance is uneven. The commercial promise of automated translation must be understood against that background, where both publishers and translators have legitimate financial concerns and relatively little margin for error.

The economic consequences may extend to the languages and authors that receive attention. Literary translators frequently discover books, establish relationships with foreign publishers, prepare samples, and persuade American editors to consider works from markets they know poorly. Their advocacy can precede a publishing agreement by years. A system that reduces the cost of producing English prose does little on its own to replicate the research, cultural knowledge, and personal relationships through which many translated books are first brought to publishers' attention. Whether automation expands the diversity of available literature will depend partly on how publishers use the resources it saves.

The Limits of Copyright

The copyright case introduces questions that employment contracts and professional advocacy cannot resolve alone. Under American law, a translation authorized by the original rights holder may receive copyright protection for the translator's original contributions. The underlying novel remains protected separately, and the translator's ability to license or reuse the translation depends on the rights retained under the relevant agreement. As AI-assisted production becomes more common, publishers and translators may face increasingly difficult questions about which parts of a finished text reflect human authorship and whether contractual descriptions of editing accurately represent the work performed.

American copyright law generally requires human authorship, although original human contributions to AI-assisted material may qualify for protection. A translator who revises a machine-generated draft could therefore possess copyrightable contributions without necessarily obtaining protection for every element of the resulting work. The extent of those rights would depend upon the particular text, the nature of the human revisions, and the contractual arrangements governing the commission. Existing law offers principles for assessing authorship, but it does not provide a universal answer for every workflow involving machine translation.

The litigation against OpenAI and Microsoft presents a separate dispute over the use of copyrighted material in training. One of the central questions concerns the fourth factor of fair use, which examines the effect of a particular use upon the potential market for, or value of, the copyrighted work. The author plaintiffs and supporting organizations argue that systems developed using protected material can damage existing licensing markets and produce commercially competitive substitutes. Their position is that the scale and commercial character of the alleged copying should weigh heavily against the defendants' claims of fair use.³

Other organizations representing authors have challenged parts of that reasoning. On October 30, Authors Alliance, a nonprofit organization concerned with access to knowledge and the public interest in copyright, discussed its own brief supporting the defendants on the narrow question of market dilution. The organization argues that copyright protects an author's expression without guaranteeing protection from every new work competing for readers. A novel may lose sales to another novel without the competing author infringing copyright, and a rule broad enough to treat all such competition as a protected market injury could extend far beyond artificial intelligence.⁸

Authors Alliance expressly declined to take a position on other disputed questions, including how OpenAI obtained its training materials and whether particular model outputs infringe copyright. Its argument concerns the legal distinction between harm caused by copying protected expression and the effects of new works that compete without reproducing protected material. The coalition supporting the Authors Guild places greater emphasis on the potential loss of markets made possible by the defendants' use of copyrighted works. Both positions address consequential questions about the economic purpose and limits of copyright, but neither determines the outcome of the litigation.⁸

A judgment in the consolidated case could influence the terms under which AI developers obtain and use copyrighted material, the importance of licensing agreements, and the kinds of market injury courts recognize. It would not automatically settle the compensation owed to translators, establish standards for machine-assisted editing, or require publishers to credit translators whose work appears in their books. Those matters will continue to depend on negotiations, professional advocacy, and the willingness of publishers to recognize translation as an independent literary contribution.

The dispute has nevertheless given translators a prominent place in a debate that has often concentrated on the rights of original authors. A novel's publication in another language creates a second relationship of authorship, involving someone whose name may appear on the cover, inside the copyright page, or nowhere prominent at all. That person's decisions determine how the original work survives the passage between languages. As publishers reconsider who will make those decisions, translators are asking the courts to recognize the economic conditions that allow their work to exist.

In her October interview, Jaquette described an experience familiar to anyone who has read literature in translation: the apparent ease of a sentence can conceal the difficulty of producing it. A reader may never know which words proved impossible to carry across, which rhythms had to be recreated, or which apparently ordinary expressions demanded hours of consideration. The translator's success often leaves little trace of the struggle. It is precisely that unobtrusive labor that publishers are now being asked to price, automate, and, in some cases, replace.



  1. Chandler, Natalie. "Literary Translators Warn AI Threatens Their Livelihoods in Copyright Case." Publishers Weekly, November 18, 2026. Reporting on the American Literary Translators Association's participation in the October 30 amicus brief and its position on copyright and professional displacement.

  2. Spratford, Sam. "Literary Translators Brace Against AI Push." Publishers Weekly, September 11, 2026. Industry reporting on AI-assisted translation, HarperCollins's automation plans, Nuanxed, and translators' concerns about professional authorship.

  3. Publishers Weekly. "Authors Guild, Co-Plaintiffs Seek Summary Judgment in OpenAI Case." September 9, 2026. Legal reporting on the consolidated copyright litigation, the parties' competing motions, and the fair use issues before the court.

  4. Publishers Weekly. "Publishers File Amicus Brief in OpenAI Copyright Case." November 2026. Reporting on the Association of American Publishers and allied organizations' intervention concerning training materials and licensing markets.

  5. Spratford, Sam. "Frankfurt Book Fair 2026: Five Questions with Elisabeth Jaquette." Publishers Weekly, October 7, 2026. Interview addressing the distinctive difficulties of literary translation, machine-generated errors, and the professional standards of Words Without Borders.

  6. Baron, Eva. "Harlequin France to Test AI-Assisted Translation." Publishers Weekly, January 5, 2026. Reporting on the HarperCollins France experiment, translators' objections, and the publisher's commercial explanation.

  7. PEN America. "Fairness in Publishing: The State of Literary Translation in the U.S.." March 31, 2026. Research into translation compensation, royalties, contractual practices, attribution, and professional conditions. See also "PEN America Finds Literary Translators Underpaid, Underappreciated," Publishers Weekly, April 2026.

  8. Hansen, Dave. "Authors Alliance Files Amicus Brief in In re OpenAI Copyright Infringement Litigation." Authors Alliance, October 30, 2026. Explanation of the organization's legal position on market dilution, competition, and the limits of copyright protection


 
 
 

Comments


FOR THE WRITERS ® AND ITS AFFILIATED MARKS ARE REGISTERED TRADEMARKS. © 2019–2025 FOR THE WRITERS.

bottom of page