Macmillan 员工说，出版商扔一个政党工作场所AI助手那后来被杀害

中文版正在随流水线自动整理，稍后刷新即可；可先阅读下方英文原文。

本文由 AI图鉴自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。

As staff at the Lower Manhattan headquarters of Macmillan Publishers headed into work one morning in January 2024, they filed past something strange on the walls—flyers welcoming a new member of the team named “NOA.” The cafeteria was festooned with purple balloons and banners, according to several workers present, and with NOA’s name in light-up marquee letters. A plate full of doughnuts spelled out “NOA IS HERE.”

NOA was an acronym for No Ordinary Assistant, which many of Macmillan’s human assistants found “offensive,” multiple employees tell WIRED. In an email to staff, NOA was introduced as an AI chatbot developed by the German software company Chaptr, which is “backed” by Holtzbrinck, Macmillan’s parent company.

The email also assured employees NOA was “a safe environment where you can input proprietary information without the risk of sharing that data with any third party.” Ostensibly, this means NOA was a closed-loop enterprise tool that did not retain or share data for model training.

Macmillan staffers say they were warned during training sessions that NOA could hallucinate and would always need to be fact-checked by humans. Still, employees say, they were encouraged to use NOA for composing emails, drafting copy, doing marketing research, and generating ideas. Multiple staffers say they were told to “only” paste a “percentage” of authors’ manuscripts into NOA. Employees could also adjust NOA’s speaking tone, with options including “lighthearted,” “spartan,” pessimistic,” “sarcastic,” and “regretful.”

One Macmillan employee remembers a Chaptr representative reassuring workers during an early in-person training session that Chaptr understood publishers were dealing with something very important—content. “And here I thought we were publishing literature,” the employee says.

Later in 2024, Macmillan employees were encouraged to use NOA to generate their goals for performance reviews. “We built a specialized prompt to have NOA write goals for you!” one email read.

Several Macmillan employees say NOA was unhelpful and unreliable. “It was terrible, and no one used it,” one says. WIRED spoke with multiple publishing employees for this article, including some who have since left their employers. WIRED is not naming these employees, nor the authors we interviewed, because they are concerned about professional repercussions.

In October 2025, Macmillan staff learned via a company-wide email that Chaptr would be sunsetting NOA. Since then, workers say, leadership has been attempting to integrate Google Gemini into the publishing process. Some staffers subsequently experimented with using Gemini to evaluate submission letters from literary agents, but the Gemini-generated evaluations were “useless,” according to one editor.

In response to WIRED’s request for comment, Kelly Collins, vice president of communications and community at Macmillan, says the publishing house uses Gemini for its “business” but “there has not been any directive from Macmillan leadership to use Gemini for any part of the publishing process, including to evaluate submission letters from agents.” She also says “the points shared do not accurately reflect how NOA was rolled out and communicated” but did not provide further elaboration when WIRED asked which aspects were inaccurate.

Collins notes that the company “takes a thoughtful approach to AI, focused on employee awareness and education while always keeping people front and center.”

Macmillan is usually cited as the smallest of the Big Five US book publishers based on the number of books sold each year. Its imprint Farrar, Straus and Giroux publishes literary fiction heavyweights like Sally Rooney, Yiyun Li, and Ben Lerner while Tor Books is a home for science fiction and fantasy authors John Scalzi, Martha Wells, and Nnedi Okorafor.

Since I first reported on the creep of AI in the publishing world, I’ve heard from staff at all of the Big Five who claim that executives are pushing them to use AI tools, even as Hachette, Macmillan, and Penguin Random House Verlagsgruppe (its German-language division) are suing some of the leading AI companies for copyright infringement.

At the same time, multiple bestselling Big Five authors say they’ve gone to “war” with their publishers over AI and are considering abandoning traditional publishing altogether.

At Penguin Random House (PRH), the largest of the Big Five after the 2013 merger between Penguin Books and Random House, several workers say they’ve been encouraged to use ChatGPT Enterprise in AI learning sessions, even though PRH’s German division is suing OpenAI for copyright infringement.

In February, representatives from Hay House—a new-age self-help publisher acquired by Penguin Random House in 2023—presented a slide deck to some PRH employees about how ChatGPT Enterprise had become a “core productivity, creativity, and collaboration tool” since they adopted it a year prior. One of the four return-on-investment pillars in the deck was “Creativity,” including “idea generation, content development, and storytelling innovations.”

Despite encouragement from the C-suite, all of the PRH workers WIRED spoke with say they’re avoiding ChatGPT whenever possible. “It makes it harder for us to do our jobs,” one employee says. Other staff say their managers have explicitly asked them not to use AI for creative tasks like publicity copy and cover design, which has caused some confusion.

“Penguin Random House approaches AI as we would any significant new technology,” says Claire von Schilling, executive vice president and director of corporate communications and social responsibility for the publisher, via email. “We are exploring ways it may appropriately support our work, guided by clear principles and safeguards, including protecting the intellectual property of our authors and the company and preserving the essential role of human creativity and judgment in publishing.”

Von Schilling also says that PRH’s “publishing groups and imprints have discretion to determine whether and how AI tools may be appropriate for particular workflows, including areas such as marketing, publicity, and design.”

Meanwhile, several bestselling Big Five authors say their relationships with publishers have suffered over the past year due to AI use. Combined, the Big Five authors WIRED spoke with have sold more than 30 million copies. One bestselling author says they received marketing copy that included AI-hallucinated quotes from their book, but the publisher insisted they were not AI-generated. “I was gaslit on all of it,” the author says. During a recent round of contract negotiations, another bestselling author says they could not get a single Big Five publisher to promise no AI would be used in any way on their books.

“It’s undermining the necessary sense of trust that has to exist between writers and publishers,” one author says.

“It’s also frustrating to watch publishers get these massive settlements [from] AI companies and then embrace AI to do their work,” another author says. “It feels like they’re feeding at both sides of the trough.”

Authors say cost-cutting layoffs at the Big Five have contributed to deteriorating conditions for their editors and forced some to embrace AI for survival. “We’re seeing editors not reading the material at all and using Claude to create marketing language that barely has any relation to the book,” one author says. Other authors and agents say they’ve been asked by editors to write their own plans for revisions “because the editors don’t have time to do it themselves.”

All of the authors I spoke with say they want publishers to communicate when and how they’re using AI. “The bare minimum we're asking for is disclosure,” one author says. Another wishes the Big Five “would get together and come up with a set of principles around AI that they intend to adhere to.” But some bestselling authors say their publishers’ use of AI is forcing them to consider publishing future books on their own.

“What are authors even doing with traditional publishers anymore if they’re going to outsource everything to a machine?” one bestselling author says. “Why can't an author do that themselves?”

Following WIRED's previous reporting, several high-profile authors condemned the adoption of AI tools in publishing. “I have never looked down on other publishing models (zine girly from way back) & every day, self pub looking ever more possible for future work,” Mo Ryan wrote on Bluesky. Her most recent book, Burn It Down: Power, Complicity, and a Call for Change in Hollywood, was published by HarperCollins.

John Scalzi, the sci-fi novelist who signed a 10-book deal with Macmillan’s Tor Publishing Group in 2024, wrote, “I would be EXTREMELY annoyed to discover my PR is ‘AI’-generated, and as someone bombarded with dozens of ‘AI’-generated emails daily, I assure you the stink of ‘AI’ is immediately noticable [sic] and immediately dismissible. Publishers are not doing their new books any favors in using it.”

Bestselling authors who have already accumulated huge readerships say they don’t need the institutional support of a Big Five publisher to sell copies. Brandon Sanderson, for instance, publishes some of his fantasy novels through Tor Books, an imprint of Macmillan, but raised more than $15 million from his readers on Kickstarter to publish four books on his own in 2022.

“Publishers have always seen authors as a bit of a nuisance, and now there’s a desire to make us obsolete,” says one bestselling Big Five author. “What’s fascinating to me is that in the process of trying to make authors obsolete by using AI, the publishers are rendering themselves obsolete.”

Auto-fetched and translated by AI Atlas. All rights belong to the original author.