Major record labels just sued Suno and Udio for copyright infringement

Udio and Suno are not the hottest new restaurants in town. They are AI startups making tools that can turn a simple text prompt into a radio-ready song, vocals and all.

And now, the music industry is coming for them.

On Monday, the Recording Industry Association of America (RIAA), representing the major record labels, filed massive copyright infringement lawsuits against both startups. The labels allege infringement on an almost unimaginable scale, claiming the companies illegally scraped millions of copyrighted sound recordings to train their AI models.

The damages they are demanding? Up to $150,000 per infringed work. Given the sheer size of modern AI training datasets, that tab could quickly run into billions of dollars.

The hits, the hype, and the receipts

Both startups are young, but they blew up almost overnight.

Suno launched in December from a Cambridge team previously behind Kensho, and immediately partnered with Microsoft to power music creation inside Copilot.

Udio dropped earlier this year, backed by Silicon Valley heavyweights like Andreessen Horowitz alongside music insiders like Will.i.am and Common. It was the engine behind King Willonius's viral Drake diss track BBL Drizzy, which producer Metro Boomin later opened up for the entire internet to remix.

To build their case, the labels included side-by-side musical notation comparisons and audio examples in the lawsuits. In one Suno generation, the track literally opens with the sung phrase Jason Derulo in the exact cadence the pop star uses in his own songs. Another Udio sample tracks closely with Green Day's American Idiot.

The twist is that the RIAA is not necessarily suing over those specific outputs. Instead, the labels are arguing that these eerie soundalikes prove the models could only have learned this by ingesting original copyrighted master recordings.

The battle over fair use

Neither Suno nor Udio has ever published their full training datasets. Udio CEO David Ding previously said they train on high-quality, publicly available music, while a Suno co-founder described their data on Discord as a mix of proprietary and public inputs.

In response to the lawsuits, Suno CEO Mikey Shulman said their technology is transformative and noted that the tool explicitly prohibits users from targeting specific artists in prompts. Udio has not publicly responded.

Paul Fakler, an IP partner at law firm Mayer Brown, pointed out that putting musical notation comparisons in the complaint is actually a bit wacky. Record labels typically own the sound recording copyrights, not the underlying musical composition rights, meaning notation similarities do not automatically prove infringement of the master tapes.

The whole case will likely boil down to fair use.

The labels argue that generating unlimited commercial soundalikes directly cannibalizes their catalog and upends the lucrative sample licensing market. But Fakler argues the startups have a legitimate defense. If the software only makes temporary copies to extract patterns and abstracts those musical traits into neural network weights, that is fundamentally no different from a human musician learning their craft by listening to the radio.

Why it matters

For creators and casual listeners, Suno and Udio have completely democratized music production. Being able to prompt a full song into existence within thirty seconds is revolutionary, but this lawsuit is a loud warning shot: any commercial track you build with these tools today carries a massive legal question mark hanging over its head.

For the broader tech sector, this lawsuit marks the moment generative AI's copyright reckoning moves from text and images directly into sound. The music labels are notoriously litigious and protect their catalogs far more aggressively than book publishers. If the courts rule that scraping master recordings cannot be protected under fair use, it will force every single audio AI model to strike costly licensing agreements or shut down entirely.

We think the labels are playing a very smart hand here. By targeting the training data rather than just the output tracks, they are aiming straight for the startups' legal jugular. Once the case moves into the discovery phase and the companies are forced to disclose what was actually scraped, Suno and Udio might find that their fair use defense is much harder to maintain in front of a jury.