Tidal's AI-label rule draws a line — but the royalty question is still open

On 29 June 2026, the music streaming service Tidal announced that tracks uploaded by artists which are wholly or substantially generated by artificial intelligence must carry an on-platform label disclosing that fact — and that such recordings will not be eligible for royalty payouts. The policy is one of the first concrete markers laid down by a major streamer in a debate that has otherwise run on industry white papers, label complaints and a steady drumbeat of confused artist posts.

The clean headline is misleading. Tidal has answered one half of the question — what disclosure looks like inside its own app — and left the harder half, the royalty treatment of AI-assisted (rather than AI-generated) music, almost entirely to the next round of negotiations. The company is choosing the visible fight, not the structural one.

What the policy actually covers

The disclosure rule, as Tidal described it on 29 June, applies to music "wholly or substantially" produced using AI tools. Songs that meet that threshold must be tagged on the platform, and royalties attached to those streams will not flow back to the uploading artist. Tidal did not, in the announcement, specify how the threshold would be enforced at upload, what audit mechanism it will use to detect undisclosed AI use, or whether human-edited or human-arranged material produced largely with generative tools would fall on one side of the line or the other.

For an artist whose work is genuinely acoustic or whose AI use is incidental (a vocal-tuning chain, a mastering plug-in), the rule changes little. For an artist working with generative vocal synthesis, lyric generation, or fully composed AI instrumentals, the rule changes the calculation: the music can still be hosted, but it will not pay.

The counter-pressure from rights-holders

The major labels have pushed, for most of 2026, in the opposite direction — for AI-generated content to share royalty pools with the human artists whose recorded work was used to train the underlying models. The structural argument from Universal Music Group, Sony Music and Warner Music Group has been that an AI vocal trained on a specific singer's catalogue is, in effect, a derivative work that should remunerate that catalogue. The labels have not, however, consistently committed to passing any such royalties back to the human session musicians, songwriters or background performers whose likenesses and recordings fed the training data in the first place.

Tidal's policy sidesteps that fight by refusing to pay anyone. A wholly AI track generates no royalty liability for the streamer on either side of the rights ledger. That is a tidy answer for Tidal — and a deferral of every question that follows.

Who carries the enforcement cost

Platforms rarely announce without a detection plan. The harder problem sits in the upload pipeline: distinguishing a wholly AI track from a heavily AI-assisted one, distinguishing a synthesised voice from a heavily processed human voice, and doing so at the volume of a streaming service's daily intake. Tidal has not published a description of the technical mechanism by which it plans to do this.

There is also an incentive problem. If Tidal's royalty ban pushes AI-native creators onto platforms with more permissive payout terms, the catalogues that thrive are on competitors' servers — and Tidal ends up the more conservative, smaller store. Conversely, if the ban holds and AI-generated tracks accumulate at scale on more permissive platforms, those platforms inherit a rights exposure the labels will eventually litigate. Tidal is, in effect, choosing to be the platform where AI provenance is recorded rather than the platform where AI catalogues earn.

What remains unresolved

The biggest open question is the assisted-work case. A song written by a human, demoed with a synthesised vocal, then re-recorded by a human vocalist with an AI-assisted mix — does it qualify as "substantially" AI? The 29 June announcement offered no test. Industry trade bodies have not yet converged on a shared definition. Until they do, the disclosure rule is more a marker than a regime.

The second open question is what happens to the catalogue already on-platform. Tidal's rule, as stated, applies to uploads; it does not address retroactive takedowns, royalty clawbacks, or audits of existing AI-generated content. A meaningful policy would need to cover each of those. Tidal will arrive at them, or the rights-holders will arrive at them first.

Desk note: this article is built around a single confirmed event — Tidal's 29 June 2026 announcement as reported by Variety — and treats the surrounding industry pressure as context rather than as further verified fact. Where the underlying technical mechanisms, retroactive treatment and AI-assisted thresholds are discussed, the sources do not specify those details, and that uncertainty has been flagged in the text rather than papered over.

Wire provenance

This editorial synthesis draws on the following public wire/social posts:

• https://en.wikipedia.org/wiki/Tidal_(service)