AI Giants Accused of the 'Largest IP Theft' in Music History

Major technology companies including Google, Microsoft, Meta, OpenAI and X have been accused of systematically scraping copyright-protected music from millions of artists to train their Gen AI systems.

The International Confederation of Music Publishers (ICMP) has compiled extensive evidence over the past two years, showing that songs by The Beatles, Mariah Carey, The Weeknd, Beyoncé, Ed Sheeran and Bob Dylan are among those allegedly used without the consent of the artists.

The trade association, which represents 90% of the world's commercially released music, has described this practice as "the largest IP theft in human history".

A systematic infringement of IP law

John Phelan, the Director General of the ICMP, says that his organisation has documented "tens of millions of works being infringed daily".

The evidence includes private datasets showing illegal scraping from YouTube by music-making applications Udio and Suno.

The ICMP also conducted research on Meta's Llama 3 language model, with the results suggesting that it was trained on copyrighted music by artists including The Weeknd, Bruno Mars, Childish Gambino and Kanye West.

Furthermore, court filings reveal that Anthropic's Claude AI system allegedly copied lyrics to hundreds of songs, including Don McLean's American Pie and Beyoncé's Halo.

This is not the first time that Claude has been at the centre of an IP controversy. Anthropic is currently fighting a class action lawsuit for scraping the works of thousands of authors while training its model.

AI systems admit they were trained on IP

Perhaps most damaging is the evidence that is coming from the AI systems themselves.

OpenAI's chatbot ChatGPT recently admitted that its Jukebox music-making application was trained on copyright-protected music by artists including The Beatles, Elton John, Madonna, Elvis Presley and Drake.

Elsewhere, Google's Gemini chatbot acknowledged it was "highly likely" that the company's MusicLM model had trained on copyright-protected music, which "could raise legal concerns" for commercial use.

Microsoft's CoPilot is the same, with the system allegedly replicating lyrics from Bob Dylan's ‘Knockin' On Heaven's Door’ and Michael Jackson's ‘Billie Jean’ in its output to users.

The legal battle

The ICMP’s revelation follows on from several lawsuits, with music rights holders suing AI companies for using their work without permission.

The world’s three major record labels – Universal, Warner Bros and Sony – have sued AI song generators Suno and Udio for allegedly using their music illegally.

Music publishers including Universal Music Publishing Group, Concord and ABKCO have also filed a suit against Amazon-backed Anthropic over copyright infringement.

In March, a California federal judge rejected publishers' preliminary bid to block Anthropic from using copyrighted lyrics, stating it remained "an open question whether training Gen AI models with copyrighted material is infringement or fair use".

Double standards

The ICMP's research highlights some glaring contradictions in the way tech companies regard the issue of data access.

“With AI and tech companies, all we hear is 'we need exceptions to build an open internet and access data, wholescale, without licenses, for our training',” John says.

“What our work on AI shows is that at the very same time, they're demanding everybody else get prior written permission before using their content.”

The organisation found that Facebook, YouTube, X, Google's Gemini, OpenAI, Suno, Udio, Microsoft and Adobe all expressly prohibit scraping of their content without prior written consent in their terms of service.

The response of the industry

For many, the outlook for the music industry seems bleak.

"These AI tools are being used in ways that will displace lyric writers and undermine existing royalty streams,” says a spokesperson from Concord, one of the companies bringing the lawsuit against the AI developers.

“Although large language model lyrics may never have the creativity of a human, LLMs trained on human lyrics coupled with their speed, scale and economy, will undermine the incentive to create new art.”

Regardless of the result of the lawsuit, it is almost certain that AI companies will become more highly regulated in the future.

The European Union's AI Act is already promising to enforce robust protections against fair use claims, requiring tech companies to respect existing copyright law and provide training transparency.

For John, the music industry's future "needs to be one of 'license or desist'".

What can be done to protect artists?

Thankfully, for those artists affected, many organisations are starting to do something about this practice. 

Elizabeth Moody, Partner and Chair of the New Media Practice at Granderson Des Rochers, is an expert in digital media licensing that works right at the nexus of AI and music.

Much of her time and energy directed towards securing a fairer deal for artists.

“Today, I work with over a dozen AI companies and have been at the forefront of the developing opportunities that AI presents to the music and entertainment companies,” she says.

“I worked on the Eleven Labs licenses with Merlin and Kobalt, the first of their kind to pay rightsholders, artists and songwriters on an ongoing basis for the right to train AI models on their music.”

Her work with Fairly Trained, a non-profit that fights for the responsible and ethical training of AI models, is something that will excite any musician that has been affected by this recent breach of IP.

“Many AI companies train on music without artists’ or rightsholders’ permission,” she says. 

“Fairly Trained certifies companies that have trained only on licensed datasets. 

“This way, users of the AI platforms will know whether or not they are working with a company that has ethically trained its product so that artists are compensated accordingly.”