Seven of the biggest names in publishing filed a lawsuit against Audible on Saturday, arguing that its recently announced AI-generated captions feature breaks copyright law in the United States.
The Amazon subsidiary debuted the service, Audible Captions, last month with a release lined up for the start of this upcoming school year (which, of course, these publishers want the courts to put a stop to). Designed in part for educational purposes, Audible Captions does exactly what its name would imply: It throws up captions on your phone or tablet corresponding to the words narrated in an audiobook.
“Why not just read the book then?” I can already hear you asking, because I sure as hell did. Audible’s advertising describes it as a tool for struggling readers or anyone who just prefers to follow along with smaller chunks of text. Clicking any single word also brings up its definition, which I, in all honesty, think is pretty nifty.
Seven major publishers do not find it nifty, however: Hachette, HarperCollins, Macmillan, Penguin Random House, Simon & Schuster, Chronicle Books, and Scholastic. They argue that this service would need to secure appropriate permissions for whatever books it captions since audio and text reproductions have separate licenses. Failing to do so would constitute copyright infringement, according to their views, since Audible Captions is essentially recreating and distributing their registered material.
And that might be a pretty open-and-shut case if things were that simple. But of course, when it comes to legislating emerging technology, things never are.
Audible argues that Audible Captions isn’t generating PDFs of complete books, and, per the company’s online statement, the service “is not and was never intended to be a book.” Instead, it relies on machine learning purportedly created using publicly available technology (Amazon’s AWS Transcribe) to transcribe text based on the last few words spoken in a recording, i.e. there’s no way to flip ahead or check previous text since nothing’s saved.
So basically the argument goes, Audible isn’t selling entire audiobooks converted to text. Instead, it’s selling software that lets customers do the converting limited to small segments at a time.
That may seem like a pedantic distinction, but it’s a big one in terms of legality. Via the Cartoon Network v. Cablevision ruling in 2008, courts established that it was totally legal for Cablevision to offer remote, cloud-based DVR services. Even though technically the shows customers recorded and played back all went through its data centre, the cable company argued the customers were the ones deciding what to record, not Cablevision, so they were technically creating the content. And that’s completely fine under copyright laws thanks to fair use doctrine, the courts decided.
It’s the same legal argument that kept VCRs and CD burners from being totally outlawed in the first place. So long as customers aren’t illicitly selling or sharing the content, it’s all good.
What I’m saying is, this all sounds a lot like what Audible is arguing, but who knows, the U.S. courts could see it completely differently. Maybe they’ll come up with entirely new rules when it comes to content created through machine learning. With how many companies are increasingly incorporating AI into their products and services, more legal grey areas are like this are bound to spring up.