AI Companies Are Raising Millions From Music They Took Without Permission (Guest Post)

Jamie Njoku-Goodwin, who holds an OBE and has served as both chief executive of U.K. Music and Director of Strategy to the U.K. Prime Minister at 10 Downing Street, has long been a vocal proponent of copyright enforcement. Early in 2023, while at the helm of U.K. Music—the umbrella organization representing British music creators, labels, venues, and companies—he cautioned that weak copyright enforcement could give rise to what he termed "music laundering." His concern was that, absent adequate safeguards, artificial intelligence advancements would allow tech giants to appropriate music without authorization, train their systems on it, and then produce polished, ostensibly original tracks whose true sources remained concealed.

Njoku-Goodwin once questioned whether comparing the situation to gangsters trafficking stolen goods was overly dramatic. Yet with three years of hindsight, he now believes his original caution may have understated the gravity of the problem. Data from Deezer reveals that AI-generated songs account for 50% of all new uploads to streaming platforms, and a staggering 85% of streams for wholly AI-generated music on that service are fraudulent. The fallout weighs heavily on the music sector, particularly on artists and creators. The laundering of illegal profits—whether cash or music—unfolds in three phases: the initial unlawful acquisition, the purification through multiple transactions designed to obscure the source, and the eventual reentry into the legitimate economy where it appears lawfully obtained.

AI music appears to be tracing that very trajectory. The first phase—appropriating music without permission—has been defended by AI companies asserting that using copyrighted material for training data is entirely lawful. That rationale, however, is increasingly collapsing. Recently unsealed court records from the New York Times copyright lawsuit against OpenAI and Microsoft show that even individuals within those organizations concede the practice amounts to theft. In 2023, Microsoft's own director of applied science, Brent Hecht, characterized the company's scale of copying as "an astonishing theft of unprecedented proportions" and called it "the largest theft of labor in human history," according to the documents. When tech employees themselves employ such language, the claim that mass copying constitutes "fair use" becomes difficult to sustain. The second phase—cleansing the work—took a new turn this month, as efforts are made to legitimize a product built on questionable foundations. Considerable attention has centered on AI music company Suno's agreements with record labels to license their content. Suno, which rose to prominence in AI music by training its models on tens of millions of recordings without authorization, has even asserted its latest version relies solely on licensed data. But Universal and Sony contend that this version was trained on outputs from Suno's earlier models, which they say were built on their recordings. In their words, Suno has "laundered" their content.

What remedies should be pursued? First, ongoing litigation is essential, and licensing deals must not be mistaken for amnesty. The message to AI companies must be unambiguous: taking content without permission and settling later is not a viable business model. Second, transparency is critical. Without knowing what content companies used for training, creators have no way to discover when their work has been stolen. Governments should mandate that AI developers disclose their models' training sources. Just as banks and financial institutions must trace the origins of funds—primarily to combat money laundering—the same principle should apply to AI companies to counter music laundering. Third, the industry must remain steadfast. It is encouraging that AI companies are beginning to discuss licensing, but such deals should not absolve the systematic harvesting of millions of works without consent, particularly since that very practice enabled these companies to dominate the market in the first place. According to SIQA, 93% of AI-generated music originates from Suno. These companies achieved their dominant market position through industrial-scale copyright infringement, and even if they are now striking deals, the past cannot simply be erased. Showing leniency to AI companies at this juncture only encourages and rewards such conduct. The growing pattern of companies disregarding copyright law to build market dominance, then using that dominance to dictate terms to everyone else, should concern us all.

Above all, the central question must remain at the forefront: in this new era of AI licensing deals, how are artists and musicians actually getting paid? In far too many discussions about AI music, creators are treated as an afterthought. It is imperative that artists and musicians be placed at the heart of these conversations. Three years ago, warnings about "music laundering" might have seemed alarmist, but today it resembles an emerging business model. Laundering only succeeds when people deliberately ignore where the illicit gains originated. Now more than ever, the music industry cannot afford to look away.