We know that AI giant OpenAI is interested in music. Witness its ongoing partnership with Spotify, which started with a collaboration on the latter’s AI ‘DJ’ feature, and recently extended into a full ChatGPT integration for people to discover music.
Readers with longer memories may also remember OpenAI’s development of two GenAI music models a few years ago: MuseNet in 2019 and Jukebox in 2020.
This was in the days before AI training and copyright was a hot topic. A good thing for OpenAI, given MuseNet’s ability to “combine styles from country to Mozart to the Beatles”, and its decision to showcase Jukebox with deepfake samples “in the style” of Elvis Presley, Katy Perry and Frank Sinatra among other artists.
You’d be foolish to take that approach in 2025, clearly. But OpenAI appears to be doing something else. Tech site The Information broke the news of OpenAI launching a new AI music project, and while its article is paywalled, Engadget reported on some of the details.
The project is exploring a tool that makes music in response to text and audio prompts, with students from New York performing arts conservatory The Juilliard School annotating music scores to help train the new model.
“The company envisions something that could be used to generate guitar accompaniment for a vocal track, for example, or to add music to videos,” noted Engadget.
Ears will be pricking up within the legal departments of music publishers at the mention of annotated scores, although OpenAI could start by focusing on out-of-copyright works. But if it wants its new model to be good, licensing a wider range of works (and ultimately recordings too) would be the logical next step.
A recent Financial Times report listing AI companies currently in licensing talks with major labels did not mention OpenAI, however. As its plans progress, the company can expect attention to be repaid to a 2023 appearance before a US Senate Judiciary Committee subcommittee by its CEO Sam Altman.
He was grilled then about Jukebox’s deepfakes – “That was a research release… not something which gets much attention or usage. It was put out to show that something’s possible” – and about OpenAI’s approach to musicians and copyright.
“We think that creators deserve control over how their creations are used, and what happens beyond the point of them releasing it into the world,” said Altman then. “We think that content creators, content owners, need to benefit from this technology… content owners, likenesses: people totally deserve control over how that’s used, and to benefit from it.”
In 2025, OpenAI’s record of giving content owners and celebrities control over their likenesses is… let’s just say it’s a work in progress.
After the release of its latest GenAI video model Sora and an app to create clips with it spurred a flood of deepfakes and videos featuring copyrighted characters, the company quickly had to promise more control – and future revenues-sharing – under pressure from rightsholders and talent agencies.
Spotify recently laid out a new template for AI music deals: partner first, build together, and license before launching products. OpenAI’s new project is its chance to follow suit and deliver on the ‘creators deserve control’ ethos espoused by its CEO. We hope it takes it.
‘ The preceding article may include information circulated by third parties ’
‘ Some details of this article were extracted from the following source musically.com ’
