This app is saving musicians by poisoning the AI so it stops stealing music

We have all heard songs and music created with artificial intelligence this past year. Apps like Suno have changed the musical landscape and millions of artists fear for their future. From this medium, an initiative has emerged that is poisoning the music repositories that AI steals from to create artificial music.

ChatGPT DOWNLOAD

As you well know, artificial intelligence systems need to be fed with large amounts of data, which often include materials protected by copyright.

Musicians now have a way to defend themselves with HarmonyCloak, a system that embeds data into songs that the human ear cannot detect, but will corrupt any AI that tries to reproduce them.

A tool designed to defend artists

Researchers from the University of Tennessee, in Knoxville, and Lehigh University have developed a new tool that could help musicians protect their work from being fed into the machine

It’s called HarmonyCloak, and it works by incorporating a new layer of noise into the music that human ears cannot detect, but AI ears cannot ignore.

This additional noise is dynamically created to blend with the specific features of any musical piece, remaining below the human auditory threshold. But the wandering AI models that listen to the music don’t know which parts to ignore, so they poison the well and ruin their recreation attempts.

Spotify DOWNLOAD

The idea is that creators can use the tool to add a layer of protection to their music before uploading it to websites or streaming services, where it could get caught in the AI’s nets. Something that is already used in image art, paintings, and pictures.

HarmonyCloak can be used in two different configurations. It can be adjusted to apply noise targeted at a specific AI model, with better results that continue to apply even after processing the track, such as compression to MP3.

Or it can generate noise that affects a series of models, so that the original creation is protected against any AI that tries to replicate it, including those that have not yet been developed.