AI Fair Use Rulings Give Music Labels New Ammunition vs Suno

Two federal judges delivered split verdicts on AI training and fair use last week, creating new tactical advantages for record labels in their high-stakes lawsuits against music generators Suno and Udio.

In Bartz v. Anthropic, Judge William Alsup ruled that training AI models on legally obtained copyrighted works constitutes fair use, while simultaneously finding that downloading millions of books from pirate sites violated copyright law. A separate decision in Kadrey v. Meta reached similar conclusions about training versus piracy.

Why it matters:

The rulings reshape legal strategy for the music industry's ongoing lawsuits against Suno and Udio, where labels seek up to $150,000 per infringed work.

Both AI music companies have admitted using copyrighted recordings to train their models, arguing their use falls under fair use protections. The new precedents suggest courts may accept training defenses while still holding companies liable for how they obtained source material.

Strategic implications for label attorneys:

• Focus litigation on acquisition methods rather than training processes
• Emphasize commercial competition with original works
• Document unauthorized downloading from streaming platforms

The legal blueprint:

Training gets protection

Judge Alsup characterized Anthropic's training as "exceedingly transformative," comparing it to "any reader aspiring to be a writer" who studies existing works to create something new. This reasoning could apply directly to music AI training.

Piracy remains actionable

However, the court drew firm boundaries around unauthorized acquisition, ruling that Anthropic's downloading of over seven million books from pirate sites was "inherently, irredeemably infringing."

Internal emails revealed Anthropic's leadership deliberately chose to "steal" books rather than pursue licensing to avoid what they called the "legal/practice/business slog."

Market harm standard

The Kadrey decision emphasized that fair use fails when AI outputs directly compete with training materials in the same marketplace. Music AI tools creating tracks in identical genres face vulnerability under this standard.

Industry response:

The Authors Guild, while disappointed in the training ruling, expressed relief that courts recognized "massive, criminal-level, unexcused ebook piracy." Music trade groups are studying both decisions for applicable strategies.

Independent artist Tony Justice filed new class-action suits against both companies in June, citing recent US Copyright Office guidance that training on expressive works "particularly when those works are used to generate substitutional outputs" likely exceeds fair use protections.

What's next:

Upcoming catalysts

Anthropic faces trial in December on piracy damages, potentially setting statutory damage precedents reaching $150,000 per work. Similar trials for Suno and Udio could follow in 2026.

Discovery advantages

Labels now have legal precedent to demand detailed records of how AI companies acquired training data, shifting burden to prove lawful sourcing.

The bottom line:

While AI companies gained breathing room on training fair use, labels secured powerful new litigation paths focused on acquisition methods and market substitution. The decisions transform music AI lawsuits from philosophical debates about learning into concrete battles over piracy and commercial competition.