Suno is trying to turn its AI music engine from a symbol of streaming’s spam problem into a more manageable participant in the music business. The question is whether labels on Suno-made songs can meaningfully contain a flood that extends far beyond one platform.

The company’s shift comes after mounting legal and commercial pressure. Suno is fighting infringement claims brought by major labels and artist bodies, while a German court recently found it had breached music-licensing rules.¹ A late-2025 breach allegation also intensified scrutiny of claims that the company scraped material from services including YouTube, Deezer and Genius for training data.¹

Against that backdrop, chief executive and co-founder Mikey Shulman has set out a cleanup plan: invisible audio watermarks on every output, fingerprinting that can help partners identify tracks, tighter rules against impersonation and deceptive audio, and a forthcoming download policy aimed at curbing mass uploads to streaming services.¹ Suno says the markers will be durable, difficult to remove and inaudible. “They are also not intended to pass judgment on whether a song is good, meaningful, or sufficiently human,” Shulman said.¹

The company is also turning to Musixmatch’s Sentinel service, which checks not only prompts and uploaded inputs but generated outputs against a catalog drawing on more than 200,000 publishers, including the three major music groups.² That is a notable attempt to address the copyright risk at the point where an AI track is actually made, rather than merely policing what users type in.

Yet the plan has conspicuous limits. Suno has not said whether it will build its own watermarking system or use a tool such as Google’s SynthID, and it has offered few specifics on download restrictions.³ More fundamentally, a broken watermark would leave past tracks unlabeled, while rival generators and open-source models could keep producing music with no marker at all.³

Suno’s pitch is transparency, not an artistic verdict: “Ultimately, we believe it should be up to artists and platforms to decide what they want to disclose,” Shulman said.⁴ For platforms drowning in synthetic uploads, however, that distinction may matter less than whether the detection tools work at scale.