Law
Copyright ownership
Determined by authorship under the law of the relevant country, not by a tool's marketing. If no human authored the protectable expression, there may be no copyright for anyone to own — including you.
Rights
Most artists ask the same two questions before letting any AI tool near their catalog: is the output protected, and who owns it. Here is what the official sources actually say, what is still unsettled, and where ANYANO deliberately narrows the question.
Before you read
This page is general information, not legal advice, and it is not a substitute for a lawyer in your jurisdiction. Copyright rules differ by country and this area is changing quickly — every factual claim below links to its primary source so you can check the current position yourself.
Question one
The answer turns on human authorship, not on whether a machine was involved.
In the United States, the Copyright Office addressed this directly in Part 2 of its Copyright and Artificial Intelligence report. Its conclusion was that existing law is adequate to handle AI: where a human contributes expressive elements to a work, those contributions can be protected, and AI can be used as an assistive tool without forfeiting protection over what the human authored. Where output lacks sufficient human authorship, it is not protectable — and the Office found that prompts alone generally do not give a person enough control over the resulting expression.
Practically, that pushes in one direction: the more of the creative decision-making is yours — the material, the arrangement, the edit, the performance — the more there is to protect. A one-line prompt into a shared model is the weakest version of that case.
Outside the US, treatment varies. The UK's guidance on ownership of copyright works and the EU's evolving AI rules start from different premises, so do not assume a single global answer.
Question two
Two different things get called ownership. Separate them before you release anything.
Law
Determined by authorship under the law of the relevant country, not by a tool's marketing. If no human authored the protectable expression, there may be no copyright for anyone to own — including you.
Contract
What a product's terms let you do with what you generate: commercial use, distribution, exclusivity, whether the provider retains a licence. This varies enormously between products and tiers, and is where most surprises live.
A service can grant you wide usage rights over output that is nonetheless not copyrightable. Both facts can be true at once, and only reading the terms plus understanding the authorship position tells you where you stand. If a release matters commercially, that is the point to involve a lawyer rather than a comparison table.
Upstream
Separate from the output: what was the model trained on?
The legality of training generative models on copyrighted material is genuinely unsettled. The US Copyright Office released a pre-publication report on generative AI training in May 2025, and litigation over training data continues in several jurisdictions. Nobody can honestly tell you today how every one of those cases resolves.
There is, however, a way to stop the question from being yours. If a model is trained only on material you own or hold permission for, the provenance of its training data is something you can actually state. That is the architecture ANYANO is built on, and the reason rights confirmation is a required step rather than a checkbox in the footer.
Distribution
Even a clear legal position does not tell you what a platform will accept.
Platforms set their own rules and change them on their own schedule. YouTube requires creators to disclose realistic altered or synthetic content and operates a removal process for synthetic content that mimics a person's voice or likeness. Streaming distributors have separately tightened rules on AI material and on how royalties are handled. Check the current policy of every platform you release to, at the time you release.
In the product
01
You confirm you own or have permission to use each file before it can be prepared for training. It is a gate in the workflow, not fine print.
02
A Sound is trained on the music you upload, and your uploads and models stay private to your account — never pooled into anyone else's model.
03
Imitating a named artist or cloning a voice without consent is outside what the product is designed for and outside its acceptable use.
What this does not do is promise a legal outcome. ANYANO makes no claim about whether any particular release of yours is protectable, licensable or clearable — that depends on your material, your contributions and your jurisdiction. What it does is keep the inputs to that question under your control.
FAQ
Train on material you own, keep it private to your account, and generate from your own catalog.
Related