If You’re Using AI For Music Generation, You Might Be Violating The EU AI Act

There are so many ways that AI is now being used almost every day in music production and creation. From reducing noise to separating stems to fixing dialog to even creating whole songs, chances are you’re using it more often then you think. That said, it’s now time to be aware that some of the things that we use it for might run afoul of a new EU AI Act law that just went into effect this week which may or may not have implications to your daily creative workflow.

What Is The EU AI Act?

The EU AI Act is the European Union’s attempt to clamp down on deepfake images, audio, or video that have been produced by or altered using AI in such a way that it’s difficult to tell the difference between a real person, place, or event from what’s authentic.

It can be boiled down to the content must be similar to something real, and capable of appearing genuine. If that’s the case, it must be clearly disclosed that AI was used in the content’s creation.

It’s early in the Act’s lifespan so it obviously hasn’t been tested yet, but it’s understood that the AI-assisted operations that we use in the course of production don’t require disclosure. That means noise reduction, stem separation, AI-assisted EQ or compression and the like don’t qualify as law breaking.

If something is generated similar to an artist, like using AI to replace a vocal phrase or guitar line that dropped out during recording, that qualifies. Although that’s not a big deal for music, it can be a problem for a dialog editor now used to replacing a line with an AI-generated voice copy instead of getting the actor back for an expensive ADR session.

Is Already Published Content Subject To The Law?

The law took effect on August 2nd, and deepfakes, audio, image, and video, anything created (that’s the key word) before that date is not subject to disclosure.

Text is a different story, where if it’s been published before August 2nd, it doesn’t need to be disclosed. In other words, it’s the publish date and not the creation date that’s important with text.

To be safe, just remember that everything AI-generated that might be deemed inauthentic is subject to the law after August 2nd.

The Penalties Are Stiff

Even if the content was created in the United States, if it’s consumed in the EU then it’s still subject to the law. And the penalties can be quite stiff.

There’s a moving scale of penalties that reach as high as 35 million euros (around $44.5 million at today’s exchange rate) or 7% of global turnover for the most serious breaches, with lower levels applying to other cases of non-compliance. Obviously this is aimed more at large companies rather than home studios.

That said, there is no AI police to enforce the law, and you can bet that even if there were, they wouldn’t come knocking on your door demanding you whip out your credit card.

In fact, the big stick that the EU has is to pressure the various platforms to take down the offending content. If that’s the case, even if it’s not costing you monetarily, it’s still costing you in time if your works suddenly disappear.

The bottom line is that this is just the opening blow in the war of deepfakes. Unless there are unforeseen circumstances that show up down the line, expect similar laws to be passed in other countries as well.

A good resource for more information on the EU AI Act for audio production can be found at Production Expert.

---