Suno hopes to go legit with watermarks for AI-generated music
Ars Technica Ryan Whitwam ● Covered by 3 sources
Suno, the AI music generator flooding Spotify with synthetic songs, says it'll start watermarking everything it makes. The move follows growing pressure to label AI content before regulators or platforms force the issue.
Suno has built its reputation, and its user base, on making full songs from a text prompt in seconds. That speed and volume is exactly why streaming services have started drowning in machine-made tracks that sound plausible enough to slip past casual listeners. Now Suno's CEO Mikey Shulman says the company wants to get ahead of the problem rather than keep contributing to it.
In a blog post, Shulman frames the coming watermarks as a response to what he calls emerging industry standards for labeling AI content, not as damage control. Every audio output from Suno's models will soon carry an embedded marker, and platforms like Spotify will get the choice to tag that content as AI-made or simply refuse to host it. That's a meaningful shift for a company whose entire pitch has been effortless music creation with few guardrails.
Shulman hasn't said whether Suno will build its own watermarking system or license one, and Google's SynthID is the obvious candidate sitting right there. Google has already opened SynthID up to outside companies, and the numbers it touts are staggering: 60,000 years of audio generated through Gemini models, more than 100 billion images and videos, all carrying invisible tags meant to survive editing and compression. If Suno goes that route, it would plug into an identification system that's rapidly becoming the default across the industry rather than inventing its own from scratch.
What's notable is the timing. Suno has faced lawsuits from major labels over training data, and watermarking doesn't resolve any of that legal exposure. But it does address a separate, more immediate mess: the sheer scale of AI tracks quietly polluting playlists and royalty pools meant for human artists. Labeling won't stop people from making AI songs, but it gives platforms a lever to sort them, and gives listeners a chance to know what they're actually hearing.
My take
Watermarks are a start, not a solution, and companies love announcing them because they signal responsibility without costing much. The real test is whether Spotify and its peers actually use the labeling to protect human artists' royalty pools, or whether this just becomes a compliance checkbox while AI tracks keep quietly outnumbering real musicians in the streaming pile.
Read more about this at: Ars Technica
