Some of the world's largest book publishers have jointly filed a lawsuit against Amazon-owned audiobook company Audible today over a new, controversial speech-to-text feature the literary industry claims is a violation of copyright law.
The lawsuit, filed in the Southern District Court of New York, includes the Big Five: Hachette, HarperCollins, Macmillan, Penguin Random House, and Simon & Schuster. It also includes San Francisco-based publisher Chronicle Books and Scholastic, the major children's publisher that owns publishing rights to Harry Potter and The Hunger Games. All seven plaintiffs are members of the Association of American Publishers.
Publishers are taking issue with Audible's new Captions feature, first unveiled last month and set to go live in September through partnerships with US public schools. The feature uses machine learning to transcribe spoken words into written ones, so users can read along while they listen to an audiobook. The issue, however, is that Audible is doing this based on audiobook recordings, which have separate licenses to physical books and ebooks. The company is not apparently obtaining the necessary licenses to reproduce the written versions of these works.
Because Audible is relying on artificial intelligence, it appears the company is trying to claim a distinction between a newly created piece of text composed using AI, based on an audio recording, and the potentially near-identical text version of the book the audiobook was created from. (As evidence that the text is generated on the fly, Amazon says its transcriptions may contain errors and are not intended to be complete recereations of the text version of a book.) At the time of its launch, Audible CEO Don Katz positioned Captions as an educational feature designed for schools, telling USA Today, "We know from years and years of work, that parents and educators, in particular, understand that an audio experience of well-composed words is really important in developing learners."
"Audible's actions — taking copyrighted works and repurposing them for its own benefit without permission — are the kind of quintessential infringement that the Copyright Act directly forbids," the complaint reads. "If Audible is not enjoined, Audible will take for itself a format of digital distribution it is not authorized to provide, devalue the market for cross-format products, and harm Publishers, authors, and the consumers who enjoy and rely on books."
"This is one of many lawsuits that will help define the future of intellectual property rights in the digital age. It raises major questions over what impact artificial intelligence, when it interacts with copyrighted material, will have on intellectual property rights," Sam P. Israel, a copyright attorney and founder of Sam P. Israel P.C., told The Verge over email. "Ultimately, unauthorized reproductions of copyrighted material, even when done unwittingly through the assistance of AI, will likely not pass muster in the courts."
The case happens to have a strong analog to a former Amazon publishing controversy a decade ago, when the company tried to launch a text-to-speech feature for its Kindle platform that would effectively do what Amazon Captions does today, but in reverse.
Publishers at the time were enraged, accusing Amazon of trying to trample on the nascent audiobook market and the licensing rights that publishers believed would help it become a thriving business. Amazon eventually caved in that regard, allowing publishers to disable the Kindle text-to-speech feature after a massive outcry from the US Authors Guild.
Short version - Audible can generate some readable text based on audio versions and publishers are mad because Audible only has rights to the audio version not a text version. Because it's an "ai" generated thing though it's not really clear to say the transcription is the same thing as the text version of the work. Video of the thing