AI Music Deals Move Forward. Musician Pay Is Still in Dispute.

Suno and Udio can turn a text prompt into a complete song with vocals, lyrics and instrumentation. The fight around them now has two fronts: whether AI companies needed permission to learn from recordings, and whether licensing deals can leave the musicians behind those recordings without a share of the proceeds.

The first front has begun to produce deals. The Recording Industry Association of America announced copyright infringement cases against Suno and Udio in June 2024, alleging unauthorized training on copyrighted recordings. Sony Music Entertainment, Universal Music Group and Warner Music Group are among the main plaintiffs. Universal settled with Udio in October 2025; Warner settled with Udio and Suno the next month; Sony’s lawsuits continue.

Training-data disclosure is a separate pressure point. Journalist Alex Reisner’s AI Watchdog project has helped musicians, including SZA, identify their work in AI training datasets. Suno, meanwhile, acknowledged in court filings that its training data included essentially all reasonably high-quality music files accessible on the open internet, alongside other data.

A license can answer one question and reopen another

Those settlements generally move toward licenses for music used in model training. Universal’s arrangement gives artists an option to opt in or out. That is a meaningful shift from the underlying allegation of unlicensed use, but it does not itself establish how payments move from an AI company or label to the singers, players and songwriters whose work helped create the licensed catalog.

The contrast matters because a fair-use outcome would address whether earlier ingestion required permission. Licensing deals, by contrast, create a commercial route forward—but leave the distribution of their value as a separate negotiation. The Music Artists Coalition’s position is not to stop AI, but to give creators meaningful control in those negotiations through consent, compensation and transparency.

The same catalog, different claimants

The American Federation of Musicians has put that distribution question into court. The union sued Universal and Warner, alleging the labels licensed recordings made by members for AI use without compensating or crediting those musicians. Universal and Warner have asked a federal court to dismiss the case, saying AI compensation is outside the existing Sound Recording Labor Agreement.

That case does not challenge whether the labels made AI agreements. It challenges whether those agreements trigger obligations to the musicians whose recordings were licensed. The court’s decision is not provided, leaving the relationship between label-level licenses and musician compensation unresolved.

What the dispute separates

• Training permission: whether copyrighted recordings can be used to build a model without a license under fair use.
• Label authority: whether a rights-holder’s license resolves the claims of the individual musicians on a recording.
• Revenue allocation: whether creators receive compensation when their work is used in training or covered by an AI deal.

The pressure is not limited to training. Streaming services divide a portion of subscription revenue across plays in a shared catalog. Critics including Delgado argue that a rapid increase in AI-generated tracks could dilute the share available to human-made music. That is an argument about platform economics, not proof that every AI track displaces a particular musician’s payment.

Suno’s scale makes the policy stakes more immediate: CEO Mikey Shulman said in February that the company had passed two million paid subscribers. Yet the current record offers no settled answer to the central financial question. Deals can establish permission; the unresolved fights will determine whether they also establish a traceable, meaningful share for the people whose music supplied the value.