Skip to content

Product3 publishers3 min readPublished Updated

Apple turns AI disclosure into a delivery requirement, and hands labels the paperwork

The tags Apple made optional in March are now mandatory for every content provider. The company that built AI detection in-house is asking suppliers to certify their own catalogues instead.

The Product Desk · Product desk

What happened

  • Apple emailed music industry partners on Thursday to announce Made With AI Labels for Apple Music, according to Billboard, which obtained the message.
  • The AI Transparency Tags Apple made optional in March are now required wherever AI created a material portion of the content.
  • Apple defined AI-platform-generated content, but set no threshold for what a material portion means in any other case.

Compiled by The Product DeskSomething wrong?How this is made

Why it matters

  • constraint Provenance now has to be captured at the point of creation and carried through delivery, and Apple Music's own chief says many labels lack the tooling to know what they are shipping.
  • exposure Providers are certifying deliveries to a platform that says it can already identify the generating model, so a wrong tag is a discrepancy someone else can see.
  • decision With no bar and no stated penalty, every distributor has to pick between over-tagging human work with a consumer-visible mark and risking an undefined sanction.
  • precedent Disclosure moves from a platform feature to a term of supply, which is the version competitors will be pressed to copy rather than the badge itself.

The tag has to come from somewhere. Apple's four disclosure categories are artwork, track, composition and music video [3], and in most release workflows those are four different suppliers, none of whom are the person doing the delivery. A rule that AI use in any of them must be disclosed [5] is a rule that it must be recorded when it happens and carried through every hand-off. That is a change to ingest schemas and to producer, designer and session agreements, which is a slower piece of work than adding a dropdown.

Oliver Schusser, who runs Apple Music, described the gap in April: "every label in the world is delivering AI, they might not know about it" [10]. He added that labels need to build their own tools to understand the scope of their AI use [11]. Apple, by his account, already has one, capable of identifying which model produced a delivery [9]. It is still relying on providers to self-disclose, on the reasoning that they are "best positioned to know how their content was created" [8], and it has not said whether submitted tags will be checked against its own detection [14]. The party with the detector is asking the party without one to sign the form.

The published numbers show what the label is actually for. More than a third of monthly uploads to Apple Music are entirely AI-generated [15], while AI music accounts for under 0.5% of listening [16] - a gap of at least 66 to one between share of supply and share of demand [24]. A badge that consumers see on tracks nobody streams is not primarily a consumer feature. It is a compliance surface bolted to the ingest pipe, which is where the volume is. Deezer, which has been detecting for over a year, catches around 60,000 AI tracks a day and said in July that most of them were fraud [20]: roughly 22 million a year at that rate [25].

Apple already enforces at that layer with money. It redistributed royalties from about 2 billion manipulated streams into its artist and label payout pool during 2025 and doubled its streaming fraud penalty fees in January [17]. Its email about tagging, though, says nothing about what happens to a provider that fails to tag, an omission flagged by Billboard, The Hollywood Reporter and Digital Music News [12].

Meanwhile the threshold is undefined. Apple has told providers what an AI-platform-generated track is, "anything that is primarily derived from a generative AI service" [6], and left "a material portion" unspecified for everything else [13]. Under an unknown sanction with an unknown bar, the cautious move is to over-tag, which puts a visible AI label on records that were mostly made by people. Spotify's approach labels the profile rather than the song and lets artists appeal [18]; Apple's email, as reported, sets out no appeal route and applies the mark to whatever the provider declared [2].

What to watch

  • Whether Apple publishes any threshold for "a material portion", or leaves each provider to guess where AI-assisted becomes AI-generated.
  • Whether Apple begins reconciling submitted tags against its in-house detection, and what the sanction is when the two disagree.
  • Whether Amazon Music, the remaining major service with no stated policy, opts for labelling or a ban.
Loading claim ledger
Loading source directory links
Loading share composer
Loading topic controls
Loading related stories