Every mix you drag into a browser-based mastering service leaves your hard drive, lands on a company server, gets processed by software you cannot inspect, and then sits in storage for a window of time set by a document you almost certainly did not read, which is why the terms of use behind cheap online mastering matter as much as the sound of the master that comes back to you.
LANDR says the important part out loud. Its terms grant the company a non-exclusive licence over what you send, while stating you are "NOT assigning or transferring us any copyright or ownership rights" on your recordings. The distance between a licence and a transfer is where all the confusion lives.
Producers skim past that wording because the price feels too small to argue about. Dirt-cheap browser mastering by the track now sits one tab away, so uploading an unfinished demo costs less than a coffee and feels about as serious. Your catalogue does not agree.
Service improvement clauses deserve a slower read. LANDR reserves the right to use your data to "test functionalities during the development phase" and to analyse what it holds to improve its services. That is normal software language, and it is also broad.
Who actually owns the master you paid a dollar for rarely gets asked out loud, yet the answer shapes every sync deal and label conversation that follows. You own the recording. What you granted is permission, and permission should end when you stop using the service.
Deletion is the part nobody tests. Find out whether closing your account wipes your uploads, whether support can remove a single track on request, and how long copies survive in backups after that. If the terms say nothing, treat the silence as an answer.
Unreleased material carries a different risk profile. A leaked single hurts far more than a lost mix, so think about which tracks belong on a third-party server at all, and which ones can wait until release week.
Chief executive Eliah Seton admitted the language was too broad, and in May 2025 the company committed to wording that promises it will not use your content to train generative models built to replicate or synthesise your voice, music, or likeness. Pressure worked. Wording changed.
LANDR went the other way and built a paid route in. Its Fair Trade AI programme is opt-in, pays a five-dollar advance on every eligible track plus a 25 percent share of net licensing revenue, and lets you leave whenever you want. Copyright stays with you.
Regulation is catching up in patches. Transparency duties for general-purpose AI models started applying under the EU AI Act on 2 August 2025, complete with a template that forces providers to publish a summary of the content they trained on. That paperwork is how you will eventually find out whose mixes went into the machine.
LANDR says the important part out loud. Its terms grant the company a non-exclusive licence over what you send, while stating you are "NOT assigning or transferring us any copyright or ownership rights" on your recordings. The distance between a licence and a transfer is where all the confusion lives.
Producers skim past that wording because the price feels too small to argue about. Dirt-cheap browser mastering by the track now sits one tab away, so uploading an unfinished demo costs less than a coffee and feels about as serious. Your catalogue does not agree.
The small print sets limits, not a copyright grab
A non-exclusive licence lets a platform copy, store, and process your audio so the service can actually run. It does not hand over your song. Read the scope, though, because some platforms bundle mastering with distribution, and distribution terms reach much further than a mastering engine ever needs to.Service improvement clauses deserve a slower read. LANDR reserves the right to use your data to "test functionalities during the development phase" and to analyse what it holds to improve its services. That is normal software language, and it is also broad.
Who actually owns the master you paid a dollar for rarely gets asked out loud, yet the answer shapes every sync deal and label conversation that follows. You own the recording. What you granted is permission, and permission should end when you stop using the service.
Retention windows decide when your files quietly vanish
Storage clocks are shorter than most people expect. HyFi keeps uploads for 30 days by default and stretches that to a year once you spend a credit, which is generous for the category, and still not a backup. Archive your own masters the day they land.Deletion is the part nobody tests. Find out whether closing your account wipes your uploads, whether support can remove a single track on request, and how long copies survive in backups after that. If the terms say nothing, treat the silence as an answer.
Unreleased material carries a different risk profile. A leaked single hurts far more than a lost mix, so think about which tracks belong on a third-party server at all, and which ones can wait until release week.
AI training clauses turned into an open fight
SoundCloud learned this the hard way. A clause added to its terms in February 2024 said your content "may be used to inform, train, develop or serve as input to artificial intelligence or machine intelligence technologies," and the backlash arrived the moment users noticed.Chief executive Eliah Seton admitted the language was too broad, and in May 2025 the company committed to wording that promises it will not use your content to train generative models built to replicate or synthesise your voice, music, or likeness. Pressure worked. Wording changed.
LANDR went the other way and built a paid route in. Its Fair Trade AI programme is opt-in, pays a five-dollar advance on every eligible track plus a 25 percent share of net licensing revenue, and lets you leave whenever you want. Copyright stays with you.
Regulation is catching up in patches. Transparency duties for general-purpose AI models started applying under the EU AI Act on 2 August 2025, complete with a template that forces providers to publish a summary of the content they trained on. That paperwork is how you will eventually find out whose mixes went into the machine.
