The music industry tackles AI abuses with limited success

The music industry is battling on multiple fronts: platforms, courts, and among lawmakers to prevent the misuse and exploitation of its content through generative artificial intelligence (AI). However, the results remain underwhelming.

Sony Music reports that it has already requested the removal of 75,000 deepfake tracks online, a number that highlights the widespread nature of the issue.

Despite claims that technology can now detect AI-generated music, many still argue that it is difficult to fully combat this issue. Even though AI-generated songs may appear realistic, they often display subtle inconsistencies in frequency, rhythm, and digital signature that aren't found in human voices, as explained by Pindrop, a company specializing in voice identification.

Platforms like YouTube and Spotify, two primary destinations for music streaming, are still hosts to fake rap songs by 2Pac or renditions of K-pop hits by artists like Ariana Grande, which they never actually performed.

"We take this issue very seriously and are working on new tools to improve AI detection," stated Sam Duboff, Spotify's regulatory policy head, in an interview with Indie Music Academy's YouTube channel. YouTube also mentioned that it is refining its technology with partners and could make announcements soon.

According to analyst Jeremy Goldman from Emarketer, "Malicious actors are ahead of the industry, which has failed to anticipate these developments and now must react." He adds that with billions of dollars at stake, YouTube will likely resolve this issue to avoid their platform becoming an "AI nightmare."

Beyond deepfakes, the music industry is increasingly concerned with the unauthorized use of its content by AI companies creating generative AI interfaces, such as Suno, Udio, and Mubert. Several major labels filed a lawsuit in June against Udio's parent company, accusing it of developing its software using copyrighted recordings to mislead listeners and divert potential paying users.

Over nine months later, no trial date has been set for this case, nor for another lawsuit targeting Suno in Massachusetts. A central point in these legal debates is the notion of fair use, which could limit the application of intellectual property laws under certain conditions.

Joseph Fishman, a law professor at Vanderbilt University, believes that there is still significant uncertainty regarding how courts will interpret these criteria. He also notes that if rulings diverge, the Supreme Court may ultimately need to make a final decision.

In the meantime, major AI companies continue to train their models on protected data, raising the question of whether the battle has already been lost.

"I'm not sure it's too late," says Fishman. "Many of these interfaces rely on copyrighted material, but new models are constantly being developed, which might have to comply with future legal decisions."

The music industry has also seen limited success on the legislative front. Several bills have been introduced in the U.S. Congress but have yet to gain traction. Some states, like Tennessee, have passed laws targeting deepfakes.

Adding to the challenge, former President Donald Trump championed deregulation, including in AI, which has created further difficulties for the industry. Several AI giants, such as Meta, argue that the government should clarify that using public data to develop AI models falls under fair use.

Should the Trump administration follow this advice, it could further tilt the balance against music professionals, although courts would likely have the final say. The UK isn't faring much better, with the Labour government opening a consultation to ease intellectual property laws to better facilitate AI development. In protest, over 1,000 artists released a silent album titled "Is This What We Want?" in February.

For Goldman, the fragmentation of the music industry continues to harm its efforts to address these AI issues effectively.