AI Music Copyright: Legal Storm Brews Over Training Data

It is being hailed as “the most promising new technology in music since the rise of streaming” — but it is also unleashing legal shockwaves that could reshape the industry. As tools powered by generative artificial intelligence make it possible to create complete songs from text prompts, courts are now being asked to settle a foundational question: did AI companies violate copyright law by training their systems on decades of recorded music?

The New York Times published an extensive FAQ on July 24 examining exactly these questions. The article zeroes in on two core legal issues: whether AI companies infringed musicians’ rights by using copyrighted recordings to train their models, and whether AI-generated content that mimics an artist’s style or voice could violate publicity laws.

How AI Music Generation Works

Companies like Suno and Udio have built platforms that generate original-sounding songs from simple text descriptions — think of it as ChatGPT for music. Users type prompts like “a soulful piano ballad in the style of Adele” and receive a complete audio track in seconds. The technology is powered by machine learning models trained on vast datasets of existing music, learning patterns of melody, harmony, rhythm, and production.

But here is where the trouble begins. Those training datasets, according to legal filings, contained tens of thousands of copyrighted songs used without permission.

The Heart of the Legal Battle: Sony v. Suno

A federal case in the U.S. District Court for the District of Massachusetts, presided over by Chief Judge F. Dennis Saylor IV, has become the first direct test of whether training AI on copyrighted audio constitutes infringement. During discovery, experts gained access to Suno’s training data and identified 61,026 specifically copyrighted recordings owned by Universal Music Group and Sony Music Entertainment. The potential damages exposure for Suno exceeds $9.1 billion.

Both Suno and Udio are also accused of using “stream-ripping” — software that bypasses technological protections on streaming platforms — to gather training material. This could add liability under Section 1201 of the Digital Millennium Copyright Act, which prohibits circumvention of copyright protection systems. A court has already denied Udio’s motion to dismiss DMCA claims.

The Fair Use Question

In 2025, two federal courts ruled that training large language models on copyrighted books constituted fair use (Bartz v. Anthropic, Kadrey v. Meta). The reasoning was that the models learned patterns rather than reproducing originals. But both judges explicitly noted the outcome would be different if AI outputs competed directly with the originals — which is precisely what AI music does.

AI-generated songs can be uploaded to streaming platforms, where they compete for listeners and royalties with the very works they were trained on. A study by CISAC and PMP Strategy projects that AI-generated music could eliminate $4.6 billion in annual artist revenue by 2028. Deezer reported that over 75,000 AI tracks were being uploaded daily to its platform as of April 2026.

Industry Divide: Settlements vs. Precedent

The major record labels have taken divergent paths. Universal Music Group settled with Udio in October 2025, and Warner Music Group settled with Suno the following month. Both are now reportedly pursuing licensing revenue from the AI companies.

Sony Music Entertainment, however, refused to settle and is pushing for a court verdict that would establish binding legal precedent. Meanwhile, independent artists — who lack the protection of major label deals — have launched their own class-action lawsuits. On June 25, 2026, the law firm Hagens Berman (which secured a $260 billion settlement in the landmark 1998 tobacco litigation) announced it was joining the independent-artist lawsuits against both Suno and Udio.

“Independent artists and producers represent the heart and soul of the music industry, and in the landscape of AI, they stand to lose the most,” said Steve Berman, co-founder of Hagens Berman.

Madonna Weighs In

The debate has also drawn sharp reactions from prominent artists. Madonna, speaking to Vogue Italia in an interview reported by Deadline, said: “Algorithms and artificial intelligence are the opposite of taking risks and to me that is the opposite of making art.”

She expressed concern that when “art is shaped too much by numbers, artists may stop taking risks,” adding that “creativity depends on instinct, failure, and freedom, not only on what platforms predict people will like.”

Broader Stakes

The legal battles extend beyond Suno and Udio. A separate $3 billion lawsuit filed by UMPG, Concord, and ABKCO targets AI company Anthropic for training on musical compositions. At least 17 major AI copyright cases are active across U.S. courts in 2026.

Adding to the pressure, TechCrunch reported that Suno suffered a data breach affecting 55.3 million users, with stolen data including names, addresses, and partial payment card numbers — a separate crisis for a company already in legal turmoil.

What to Watch For

Summary-judgment hearings are expected in Sony v. Suno by late July 2026. Beyond that, the Munich Regional Court in Germany is expected to deliver a verdict in the GEMA case on July 31, and the European Court of Justice is due to issue an opinion in Like Company v. Google Ireland on September 3.

The outcomes of these cases could define the boundaries of AI music for a generation. At stake is not just the future of companies like Suno — valued at $5.4 billion after its June 2026 Series D fundraising — but the fundamental question of whether the law will treat AI training as inspiration or theft.

As the New York Times FAQ frames it: the technology may be the most promising innovation since streaming, but the legal questions it raises are only beginning to be answered.