Suno will watermark every track its models generate and cap downloads across its service, CEO and co-founder Mikey Shulman announced in a company blog post. The changes arrive as Suno fights copyright suits from Universal and Sony in the United States, a German court ruling that found it violated music licensing law, and a Massachusetts suit tied to a late 2025 hack that exposed how the company trained its models. Shulman framed the shift as an effort to align with what he called industry standards for labeling AI content — a tonal pivot for a company that has become one of the largest sources of AI-generated music flooding Spotify and other streaming platforms.
Every audio output from Suno's models will carry an invisible signature embedded in the waveform, and a detector trained on that signature will let partner platforms mark tracks as AI-generated or block them outright. Shulman did not say whether Suno will build the system in-house or license Gemini maker Google's SynthID, which Google recently opened up to outside companies. SynthID has already labeled 60,000 years' worth of Gemini-generated audio and more than 100 billion images and videos, giving it the broadest deployment footprint of any watermarking scheme in production.
“emerging industry standards”— Mikey Shulman, Suno CEO and co-founder
Shulman said the watermarking technology is durable and resistant to tampering without degrading audio quality. He did not address the obvious failure mode: if the signature is ever broken, every track produced up to that point loses its AI label retroactively. And Suno's watermarks only cover Suno's output. Rival generators without watermarking commitments, plus open-weight music models that anyone can run locally, keep pumping out unlabeled AI audio regardless of what Suno does.
Key facts
- 01Suno will watermark every audio output from its models, letting partner platforms flag or block AI-generated tracks.
- 02Google's SynthID, a candidate watermarking system, has labeled 60,000 years of Gemini audio and over 100 billion images and videos.
- 03Suno faces US copyright suits from Universal and Sony, plus a German court ruling that it violated music licensing law.
- 04A late 2025 hack revealed Suno scraped training data from YouTube and Deezer, triggering a Massachusetts suit over undisclosed user data exposure.
- 05Download limits stem from a prior settlement with Warner Music Group; CEO Mikey Shulman says most users won't be affected.
The legal pressure explains the timing. Universal and Sony sued Suno in the US last year over copyright infringement, and both cases could result in significant damages if the labels prevail. A German court has separately ruled Suno violated the country's music licensing regime, opening a second front in Europe. Suno also settled with Warner Music Group in a deal that required it to limit downloads — the specifics of which the company is still working out.
Then there's the late 2025 hack. The breach revealed that Suno had scraped content from YouTube, Deezer, and other platforms to train its models — a fact the company had not disclosed. The intrusion may also have exposed private information for millions of users, and Suno did not notify them. The Massachusetts suit alleges the company should have. Taken together, the four legal actions cover training data provenance, output infringement, licensing compliance, and user data handling — most of the major exposure surfaces for a generative AI music company.
Suno is repositioning as a tool for musicians and personal projects rather than a spam factory for streaming platforms. Shulman used the blog post to emphasize the human side of music-making and to describe existing guardrails: users cannot include specific artists or song titles in prompts, and Suno partners with Musixmatch to keep copyrighted material out of user-supplied audio inputs. The company also tightened its usage policy in the same update.
“can't replace the human experiences, emotions, and imperfections that make music meaningful”— Mikey Shulman, Suno CEO and co-founder
Download limits will target what Shulman described as large-scale abuse — the pattern where users generate huge batches of AI tracks and upload them to Spotify and other platforms to farm royalty payments. Shulman said most users will not notice the caps, though Suno has not published the numeric thresholds.
The watermarking commitment matters for the wider AI music market because Suno is the volume leader. If Spotify, Apple Music, and Deezer integrate a detector — either SynthID or a Suno-native equivalent — they can filter or label AI tracks at ingest, which changes the economics of royalty-farming operations. That is roughly the outcome the major labels have been pushing for through litigation, so a functioning detector could de-escalate at least parts of the copyright fight.
The caveats are real. Watermarks are only as strong as their weakest implementation, and any leaked or reverse-engineered detector code lets bad actors verify when they have successfully stripped a signature. Open-source music models — of which there are already several — do not have to cooperate with any of this. And none of Suno's compliance moves address the underlying training-data question that the Universal, Sony, and German cases turn on: whether the company had the right to learn from the copyrighted recordings it ingested in the first place.
Suno's calculation looks straightforward: signal enough good-faith compliance to survive the current wave of litigation, cut deals with the labels where possible — the Warner settlement is the template — and hope watermarking plus download caps are enough to keep the platform viable while the courts sort out training-data liability. For the AI music market, the more interesting question is whether the major streaming services adopt the detector broadly. If they do, AI-generated tracks become a labeled category rather than an invisible flood, and the fight shifts from whether AI music exists on Spotify to how it gets paid. That is a fight Suno would rather have than the one it is having now.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
