Suno adds watermarking and download controls amid copyright lawsuits

Why it matters

Shulman is trying to make Suno output legible to the rest of the music business. Effective provenance could curb streaming fraud, but it cannot settle claims over the recordings used to train Suno's models.

Mikey Shulman is adding a new provenance layer to Suno's AI-generated songs, with audio watermarking, fingerprinting and tighter download rules intended to help streaming services identify tracks made on the platform and curb royalty manipulation.

Suno said in an August 6th blog post that it has started rolling out transparency tools and will adopt the new watermarking and fingerprinting technology over the coming weeks. TechCrunch reported that Suno has not identified the technology behind the system or provided technical specifications.

The announcement puts Shulman, Suno's co-founder and CEO, in the role he has been moving toward since Suno's copyright fights began: building the safeguards that music distributors and rights holders need without retreating from his argument that generative music should be broadly accessible. Shulman earned a physics Ph.D. from Harvard and was the first machine-learning hire at Kensho, where he later led machine learning. He has also taught natural-language processing for finance at MIT Sloan.

Shulman built Suno with former Kensho colleagues Martin Camacho, Georg Kucsko and Keenan Freyberg. Their Cambridge, Massachusetts company moved quickly from text-to-song generation into editing, voice tools, stem extraction and a browser-based production workspace. That expansion has made provenance a distribution problem. Paid users receive commercial-use rights for songs created under their subscriptions, giving them a direct route from Suno into streaming catalogs and other revenue-generating channels.

A second generation of watermarking

Suno says the coming marks will be inaudible, durable and resistant to tampering. The watermark and fingerprint are supposed to remain detectable when a song leaves Suno, allowing distribution platforms to determine where it was generated. Suno has not published test results, named a detection standard or explained how the marks will survive compression, remixing, stem separation and other common audio transformations.

This is not Suno's first watermarking claim. In March 2024, when Suno released its v3 model, it said it had developed proprietary, inaudible watermarking capable of detecting whether a song was created with Suno. The August 6th announcement describes new technology aligned with emerging industry standards and designed for closer cooperation with distribution platforms.

That distinction matters. The earlier system appears to have focused on Suno's ability to recognize its own output. Shulman's new plan treats provenance as infrastructure that outside platforms can use to police fraud and enforce their own disclosure policies. Suno has not explained whether the old and new systems will operate together or whether previously generated tracks will receive the new marks.

Suno also plans a downloads policy intended to make mass distribution to streaming services harder. The policy's limits, enforcement method and treatment of existing songs remain unspecified. Any practical effect will depend on whether Suno can distinguish a musician exporting a catalog from an operator producing tracks at industrial scale.

The updated community guidelines prohibit fake engagement, bots, attempts to evade Suno's safeguards, deceptive audio presented as authentic and unauthorized use of a real person's voice or likeness. Suno also says it works with Audible Magic and Musixmatch to screen uploaded audio and lyrics for possible unauthorized use.

The safeguards arrive under legal pressure

The timing is clear. On July 31st, a Munich court ruled that Suno infringed copyrights controlled by German collecting society GEMA. The ruling followed GEMA's allegation that Suno used protected recordings for training and could reproduce recognizable material from represented songs.

In the United States, Universal Music Group and Sony Music Group continue to pursue claims that Suno copied protected recordings to train its models. Warner Music Group, an original plaintiff in the record-label case, settled with Suno in November 2025 and entered a partnership covering licensed models and artist controls.

Suno's own disclosures establish the central training question. In a 2024 response to the record labels, Shulman said Suno trained on music found on the open internet, including copyrighted material, and argued that the process qualified as fair use. A later California training-data disclosure said Suno's models use tens of millions of public music files and associated metadata, including material that may be protected by intellectual-property rights.

Watermarking addresses what happens after Suno creates a song. It can help identify synthetic tracks, support disclosure and give streaming platforms another tool against automated catalog flooding. It does not decide whether Suno had permission to use recordings when building its models, which is the question at the center of the copyright cases.

The provenance push also follows scrutiny of how Suno assembled that training corpus. RuntimeWire reported in July that leaked code described scraping from YouTube, Deezer and Genius. Suno has said its collection methods respected access barriers such as paywalls and password protections.

Shulman's larger bet

Shulman has enough capital to turn these policies into distribution infrastructure. On June 3rd, Suno said it raised more than $400 million in a Series D led by Bond Capital at a $5.4 billion post-money valuation. IVP, Forerunner, Union Square Ventures, Alkeon and Quiet joined the round, alongside existing investors Matrix, Lightspeed, Menlo Ventures and Schroders Capital.

The financing announcement said Suno planned to release its first model developed with the music industry. The new watermarking and fingerprinting layer can support that transition by giving labels and distributors a way to identify Suno output after it moves beyond Suno's own application.

Shulman wrote that artists and platforms should decide what they disclose. That position preserves creative flexibility for users while shifting part of the enforcement burden to distributors. It will work only if those platforms can reliably detect the marks and trust Suno's implementation.

Suno's next proof point is therefore technical. A watermark that survives ordinary editing and reaches major distribution systems could help Shulman connect consumer-scale generation with licensed music infrastructure. Until Suno publishes specifications or platform integrations, its durability and reach remain company assertions.