How to License Your Music for AI Training | Elacity
Opting your music into AI training takes one click. Here are five questions to answer first: what you must own, what you grant, who counts the use, what leaving undoes, where terms live.
How to License Your Music for AI Training: Read This Before You Opt In
To license your music for AI training, you opt in through a distributor or licensing platform that has AI partners, and only for tracks where you own every right. Before you click, know five things: what you must own, what you are granting, who counts the use, what leaving undoes, and where your terms live after the deal.
Your catalogue may already have trained someone's model without a word to you. The opt-in programs arriving now are the first time you get asked, and the terms you accept in one click can outlast your ability to change them.
The newest arrived on November 18. AI audio startup Sonilo signed a licensing deal with independent distributor Symphonic to train its music model on recordings from artists and labels who opt in. Financial terms were not disclosed, and neither company said how many have opted in.
It joins a growing shelf: Ditto's AI music licensing program, LANDR's Fair Trade AI, and SourceAudio's dataset marketplace, which already invited Symphonic artists to opt in under a February deal. This guide, part of our Creator Economy series, covers what to check whichever door you choose.
Question 1: Do You Own Everything You Are Licensing?
Every program we read draws the same line. Ditto opens its program only to rights holders who own and control all of the rights to their recordings. LANDR requires you to own the recording and the publishing outright, so an uncleared sample, an unsettled co-writer or an old label deal takes a track off the table.
Sort your catalogue before you opt in, not after. A track you cannot fully clear is a track you cannot safely license.
Question 2: What Exactly Are You Granting?
Opting in is permission, not a sale. LANDR describes it as letting it place your tracks in licensed datasets that train AI audio models, while ownership of the music stays with you.
Programs differ in how far that permission reaches. Ditto lets you opt in a full back catalogue and future releases, or single releases as you upload them. BandLab says an expression of interest in AI training licensing grants it no rights to license your song for training at all.
Opting in also does not mean you get used. Ditto says inclusion is at its partners' discretion. Ask which partners, which uses, and for how long. Training a model to score videos is a different use from cloning a voice, and your consent should say which one you gave.
Question 3: Who Counts the Use?
Payment follows a count. Sonilo says it will provide attribution reporting for rights holders, building on a partnership in which Musical AI measures how licensed recordings and compositions contribute to the audio Sonilo generates.
That is real progress, and it is still a report produced on the buyer's side of the deal. We explained why whoever holds the meter decides what you get paid. The practical step for you: ask to see a sample usage report before you opt in, and ask how often it arrives.
Question 4: What Does Leaving Undo?
Every program here lets you leave. None lets you take back what a model already learned. Ditto states that if a partner trained on your music before you opt out, those models may continue to operate. LANDR compares it to a covers band that can stop learning your songs but cannot forget the one it already plays.
So treat opting in as permanent for anything already delivered, and opting out as protection for the future only. Our guide to where opt-out stops maps the same edge for writing and images.
Question 5: Where Do Your Terms Live After the Deal?
Few programs answer this one. Your terms live in a distributor's dashboard and in contracts between companies. The audio file itself carries none of them. Once it sits inside a dataset, your rules stay behind in the paperwork.
That gap is what we are building Elacity dDRM to close. In the design, you wrap a track, a stem pack or a whole catalogue into a Wealth Capsule: a programmable, royalty-bearing good with your terms written into it, so the terms travel with the work instead of waiting in someone else's dashboard. Elacity dDRM is designed so each use is checked against those terms before it happens, rather than reconciled from a report afterwards.
Elacity dDRM is not live yet, and we will keep saying so plainly until it is. Until then, the five questions above are your best protection.
A Checklist Before You Opt In
- Own the recording and the publishing, with every sample and co-writer cleared.
- Check whether the opt-in covers your back catalogue, future releases, or single releases.
- Ask which AI partners and which uses the permission covers.
- Ask how use is measured, and ask to see the report.
- Assume anything already trained stays trained.
- Keep your own dated record of what you opted in and when.
FAQ
Can I license my music for AI training if I am signed to a label?
Usually not on your own. The programs above require you to own or control all the rights, so whoever holds your masters holds that decision too.
Will opting in guarantee my music is used?
No. Opting in makes tracks eligible. The AI partners, or the program curating each dataset, choose what goes in.
Can I opt out later?
Yes, from future datasets. A model already trained on your music may keep operating after you leave.
Sasha Mitchell, Elacity's founder, puts the reason we exist plainly: "The people who create the value should own it. That is the entire reason Elacity exists."
Opt in with your eyes open, and keep your terms as close to the work as you can. Follow Elacity on X for the next guide.