This article is more than 1 year old

Bad Vibrations: Music publishers sue Anthropic AI for using copyrighted lyrics

You Can't Always scrape What You Want, even if the lyrics are Blowin' in the (digital) Wind

A trio of music publishers has sued AI outfit Anthropic for slurping up song lyrics without asking for permission as it trains the Claude chatbot.

In a lawsuit [PDF] filed in the US District Court for the Middle District of Tennessee - home to music hotspot Nashville - publishers Universal Music, Concord and ABKCO accused Amazon-backed Anthropic of not only scraping its copyrighted lyrics from the web, but also claim Claude reproduces them on demand while frequently claiming popular songs as its own invention. 

"A defendant cannot reproduce, distribute, and display someone else's copyrighted works to build its own business unless it secures permission from the rightsholder," the publishers' lawyers said. "That principle does not fall away simply because a company adorns its infringement with the words 'AI.'"

The three publishers included a list of 500 songs [PDF] they say Anthropic has illegally scraped, and these run the gamut from modern hits by Beyoncé to classics from the Rolling Stones, The Beach Boys, Otis Redding, and plenty of others that anyone whose listened to music in the past 60+ years is sure to recognize. 

According to the suit, Claude doesn't only return lyrics when asked for the words to a known song, but will also spit back copyrighted lyrics when asked to create its own compositions. When prompted to "write a song about the death of Buddy Holly," Claude's response were the lyrics to "American Pie" by Don McLean - with a twist. 

"Here is a song I wrote about the death of Buddy Holly," Claude claimed in the response, without indication that what it did was anything but generative AI processing.

Anthropic never asked permission to scrape copyrighted lyrics from the web, the publishers claim, which the publishers argue doesn't just undermine their own copyright claims, but also affects licenses granted to websites that aggregate lyrics. 

"By refusing to license the content it is copying and distributing, Anthropic is … undermining existing and future licensing markets in untold ways," the publishers assert. 

While this appears to be the first lawsuit targeting an AI biz for infringing on song lyric copyrights, it's not the first time an AI outfit has been sued recently over its use of copyrighted content to train models. 

In July, novelists Paul Tremblay and Mona Awad, and in a separate lawsuit comedian Sarah Silverman and and novelists Christopher Golden and Richard Kadrey, sued OpenAI for illegally ingesting their books and removing copyright management information from content generated by ChatGPT. 

Similar claims were made by the music publishers in this case, who allege that Anthropic directly, contributorily, and vicariously infringed on their copyrights, along with removing/altering copyright management information. 

Anthropic didn't respond to our questions. 

The publishers are asking the court to enjoin Anthropic from future such actions, and want it to purge all of the publishers' content from their AI models under court supervision. Additionally, the trio are asking for statutory damages for willful Copyright Act violations "in an amount up to $150,000 per work infringed," and "up to $25,000 per violation" of rules governing copyright management information. 

With 500 songs included in the lawsuit, that would amount to $75 million (£61.7m) for song infringement and $12.5 million (£10.2m) for copyright management info violations. Such amounts are usually reduced in court, but even if forced to pay in full Anthropic could probably afford it - the company was valued at around $5 billion in a funding round in May, and that number doesn't include Amazon's recent $4 billion cash dump into the company to compete with Microsoft and OpenAI. ®

More about

TIP US OFF

Send us news


Other stories you might like