AI Use and Output Notice

01How AI works in DAWN

DAWN contains no AI model of its own. It is an interface and an execution layer: it gathers context from your session, sends it to a model run by a provider you chose, and carries out what comes back.

This means the intelligence you experience is the provider’s. Two people running DAWN with different providers, or different models from the same provider, will get materially different results. So will the same person after a provider updates a model.

It also means we cannot fix a model’s reasoning. We can change how DAWN asks and what context it supplies; we cannot change what the model knows or how it decides.

02It will sometimes be wrong

AI models produce plausible output, which is not the same as correct output. Expect all of the following at some point:

• Confident musical or technical advice that is simply wrong.
• Invented facts — a plugin parameter that does not exist, a device that was never made, a piece of theory that is not right.
• Misreading what you asked for, and doing something adjacent to it.
• Inconsistency — a different answer to the same question, minutes apart.
• Actions in your session that are not what you had in mind.

Treat DAWN as a fast, capable assistant whose work you check, not as an authority. Your ears remain the final judgement on anything musical.

03Who owns the output

We claim nothing. We take no ownership of and no licence over anything DAWN helps you make.

Beyond that, the position is genuinely unsettled, and we would rather say so than give you false comfort:

• Whether AI-generated material attracts copyright, and who owns it if it does, is not settled in UK law or in most other jurisdictions.
• Your rights in generated audio are also affected by your provider’s terms, which may claim, disclaim, or condition ownership. Read them.
• Where you have made substantial creative choices — arrangement, selection, editing, performance — your contribution is more likely to be protected. Where output is used unedited, it is less clear.

If ownership matters commercially for a specific release — a sync placement, a label deal, a competition entry — take proper advice on that release rather than relying on a general statement here.

04Originality and infringement

AI models are trained on existing material. Generated output can resemble an existing work, sometimes closely, without you asking it to and without any warning.

We do not check output for similarity to existing recordings, and no such check is built into DAWN. Neither, in general, do the providers.

Before you release anything made with generative assistance, listen to it critically for resemblance to material you know. If something sounds familiar, it may be. Under the Acceptable Use Policy you must not deliberately reproduce existing works or imitate a real person’s voice without consent.

05What is sent, and what providers do with it

To answer you, DAWN sends your provider your instructions, session context including track names, and — if Mac control is active — a screenshot of your screen. The full detail is in the Privacy Policy.

Whether your provider retains that data, for how long, and whether it uses it to train future models, is decided by them under your agreement with them. Some providers train on consumer-tier input by default and not on API input; some offer opt-outs; some differ by region. We cannot answer this for you, and we would be guessing if we tried.

If you are working on something confidential, check your provider’s data policy before you connect it, not after.

06The agent acts on its own for small things

DAWN is not only a chat window. It carries out low-risk actions in your session without asking, and stops to confirm before consequential ones. That is a deliberate design decision — an agent that asked permission for every click would be useless.

The consequence is that DAWN can change your project in ways you did not individually approve. It keeps a redacted record of what it did, and can undo where the DAW supports it, but not everything is reversible.

Save often, and keep backups. See the DAWN EULA.

07Where responsibility sits

You are responsible for what you release. Not the model, not the provider, and not us.

That means checking output for accuracy, for similarity to existing work, and for whether it is actually good. It also means complying with the rules of any platform, label, distributor, or competition you submit to — a growing number of which require AI-assisted work to be declared.

We provide the tool. What you do with it, and what you put your name to, is yours.

Questions: support@maryalabs.io.