What exactly did Tidal change?
On 29 June, Tidal published a new AI policy with a hard money line attached. From 15 July, any track the service judges to be wholly AI-generated will earn no royalties at all, will not be eligible for Tidal's direct-to-fan sales, and will carry an AI badge so listeners can see what they are playing. The track is not deleted. Tidal is explicit that it will keep hosting AI music and let subscribers decide what they want in their ears, it just will not pay for it. From mid-July the service starts identifying and tagging the 100 percent synthetic uploads, and it says that once detection improves it will extend the badge to music that is substantially AI as well.
The rule reaches into Tidal Upload, the open channel where independent artists put their own catalogue and where most of the synthetic flood arrives. AI tracks built to deceive, the ones cloning a real artist's voice or name to skim plays, get the harder treatment: removal, using automated detection tools. Tony Gervino, Tidal's executive vice president and editor-in-chief, framed it as protecting the people who actually make the music: "We are committed to protecting and rewarding organic creativity to avoid compromising an artist's ability to connect with and build their fandom from Tidal subscribers." Tidal, majority-owned by Jack Dorsey's Block since 2021 and part-held by Jay-Z and a group of artists, has sold itself for years as the hi-fi, artist-first option, so drawing the sharpest anti-AI line of any major service is on brand.
Why should a house producer who has never opened Tidal care?
Because of how the pot works. Streaming services pay from a single pro-rata pool: all the subscription money goes into one bucket, and your slice is your share of total plays. Every junk stream that is not yours, including the AI tracks farmed purely to harvest plays, shrinks the per-stream rate that every independent producer splits. The scale is not small. Deezer reported that daily AI-track submissions climbed from about 10,000 in early 2025 to more than 60,000 by March 2026, roughly 85 percent of them flagged as fraudulent. That is a tide of synthetic audio pouring into the same pool a Lagos deep-house producer or a Leipzig minimal label draws from.
Cutting wholly AI tracks off the royalty pool, in theory, hands that money back to humans. For a working underground producer living on margins a major-label act would laugh at, even a small correction to the per-stream rate is real. This is the first time a major service has said the quiet part out loud: not every upload deserves to be paid.
The fight over AI music has finally moved from taste to money, and money is where it was always going to be decided.
Where is the catch?
In the line itself. Wholly AI-generated sounds clean until you try to draw it. A large share of underground music in 2026 is already AI-assisted somewhere in the chain: stem separation, mastering, sound design, a generated pad buried under live drums. Tidal's promise to extend the badge to substantially AI music once detection improves is exactly where a bedroom producer using cheap AI tools to compete could get caught in a net meant for the slop farms. The badge is also a judgment call, made by the platform, on detection nobody outside Tidal can audit.
Tidal is not alone, and the field does not agree on where the line sits. Spotify, Apple Music, Deezer and Qobuz have all started labelling or curbing AI uploads, each with its own definition. Tidal's no-royalty stance is the most aggressive so far, and it admits the policy is a living document, leaving open whether licensed AI models, the kind the major labels are now signing deals with, might eventually get paid after all. The principle is right. The next year of edge cases will decide whether it protects artists or just reshuffles who the platform is willing to call one.
