Suno will stamp every track its AI makes (and clamp down on mass downloads)
The watermark is inaudible, sits inside the waveform itself and is meant to let platforms recognise a Suno-made track. It lands with record labels suing the company and Europe's AI transparency rules already in force.
Suno said on 6 August that it will start marking the tracks its AI produces. The watermark is embedded in the audio, nobody hears it, and the point is that a partner platform can look at a song and tell it came from there.
The announcement came in a post signed by co-founder and chief executive Mikey Shulman, which describes the mark as durable and resistant to tampering without affecting the listening experience. One line is worth holding onto: the watermark is not meant to judge whether a song is good, meaningful or sufficiently human. It only says where it came from.
The watermark is not the whole announcement
Three things arrive together. Alongside watermarking and fingerprinting, the community guidelines gain tiered enforcement — warnings, temporary suspensions, permanent bans and, in the worst cases, referral to the authorities. And the download policy changes, to stop people pulling tracks down by the hundred.
That last one gives away the real problem. The scheme irritating everyone is simple and industrial: generate thousands of songs, push them onto streaming services, then have bots play them on repeat to harvest royalties. Suno also announced a partnership with Musixmatch to integrate its Sentinel copyright detection system.
What was left unsaid
There is no firm timeline beyond “coming weeks”, and no confirmation of which technology will be used, whether something built in house or an existing system like Google’s SynthID. None of it answers the question sitting in court, either: what music the model was trained on. The company was already under pressure when it emerged that training appeared to involve millions of tracks scraped from YouTube.
For Europe the timing is not innocent. The AI Act’s transparency rules took effect on 2 August and require AI-generated content to be marked in a machine-readable way. A durable audio watermark is precisely how you do that.
By Oliver Grant
Image: Suno
