Why Tidal cannot pay what it cannot verify

A platform just tied royalties to proof of authorship. Proof is the part no-one has launched yet. We have built it, and it will launch soon with notarized and CopyrightID time stamps…

Ask an AI assistant to recommend a song, and it will only tell you what it can confirm is true. If the facts are missing or impossible to verify, leave them out. Sometimes it does worse, filling the gap with something that sounds right but isn’t. That single behavior, confirm or omit, is quietly deciding which artists get discovered and which stay invisible, not because they lack talent, but because the facts about them live nowhere a machine can check.

Tidal just applied the same logic to money.

What changed

Spotify moderates for spam. YouTube labels AI content. Deezer runs detection on uploads. Tidal did something narrower. Starting July 15, tracks that it identifies as wholly AI-generated remain on the platform and carry a label. They will not earn royalties or direct-to-fan sales.

Uploads that impersonate artists, mislead listeners, or enable fraud are blocked outright, the same as anywhere else. Everything else, AI-made but not fraudulent, is tolerated and unpaid.

This separates two questions most platforms still bundle together: Does this belong here, and does this deserve payment?

A track can pass the first test and fail the second. But that distinction only holds if Tidal can actually tell the difference, for every track, at scale.

A question no platform can answer alone

To run this policy, Tidal needs four answers for every track:

1. Who made it,
2. Whether a human was involved,
3. Whether the work is fully AI-generated or AI-assisted, and
4. Whether the distributor disclosed that accurately.

None of these come from listening to the audio. A generative model can now produce output no ear can distinguish from a human performance. The facts that matter live upstream, in the production chain, where Tidal has no visibility, and the distributor has every incentive to describe its own upload favorably.

Self-reported status is not a fact. It is a claim from one party with a financial reason to shade it.

A record built to prove, not assert

No platform can verify any of that on its own. It needs a record that exists outside the distributor’s control, timestamped at the point of creation, that proves what a track is rather than asserting it.

That is the same gap that already decides whether an AI assistant will recommend an artist at all.

An assistant checking who recorded a song needs an independent record it can point to and say, ‘This is real; it was logged at this time, and I can prove it.’

A platform that checks whether a track qualifies for payment needs the same thing. Metadata a distributor can edit after upload doesn’t settle anything. It only shows what the uploader currently wants to be true.

A record fixed before the payment question existed is different in kind, because nobody had a reason to lie yet when it was written.

Why does timing decide who is believed?

An early record has a property no amount of money can buy back. A date written to a public ledger cannot be moved backward. When a distributor’s claim and an independent record disagree, the deciding question is simple:

Which was made first, and can that be confirmed without trusting either side? A record anchored months or years ago answers it.

A distributor asserting authorship today cannot answer it for the past, because the past is already written. Independence matters here too.

If a platform built its own verification system, it would be both the keeper of the record and the party deciding what to pay on it. That is a conflict any skeptical system should discount.

A record anchored outside every platform, checkable by anyone without requiring trust in whoever made it, does not carry that problem.

Where this goes

Tidal is one platform solving one payment dispute. But the mechanism does not stay contained. Once a major platform ties royalties to provable authorship, competitors face pressure to close the same gap or become the place unverifiable AI content routes to for a payday it cannot get elsewhere. Distributors face the same pressure in reverse: move from self-reported claims to claims they can back up, or watch the tracks they submit default to unpaid.

The fight is not over whether AI music should exist. It is over who can prove what a track is, before the platform ever has to ask. Artists and catalogs with a record made at the moment of creation, anchored outside any single company, get named, recommended, and now paid without friction.

Everyone else is arguing after the fact, trying to prove something that should have been settled before the dispute began.