Google/Universal Music talks hint at blueprint for IP rights in AI age | WARC | The Feed
The Feed
Daily effectiveness insights, curated by WARC’s editors.
You didn’t return any results. Please clear your filters.

Google/Universal Music talks hint at blueprint for IP rights in AI age
Google and Universal Music are in discussions about the licensing of melodies and voices from the labels’ back catalogue for use in artificial intelligence, in an early test of how companies (and eventually brands) will manage intellectual property in the generative AI era.
Why IP rights matter
In a world in which anyone can make anything, especially when it concerns music, literature, journalism, or film, the problem of sustaining a professional industry in all of these areas becomes difficult to ignore.
For brands, this presents as much an opportunity as a concern: a remunerative model for AI integration suggests ways in which limits might be placed on the use of intellectual property, not least on the misuse of brands’ assets and logos. But, with open source systems also available, significant questions remain.
Signs of a deal
With the genie already out of the bottle, rights holders and technology firms are looking at how they can work together to legitimately monetise some of the AI magic.
- Four people familiar with the matter confirmed the talks to the Financial Times, which come as the music industry countenances the existential threat of defending its intellectual property, as generative and deep learning technologies come for music.
- Discussions remain in an early stage, but the anticipated outcome will be a tool for users to legally create music with AI using voices and melodies from real artists in a way that will pay the copyright owners for its use. Another source tells the paper that Warner Music has also talked to Google about a similar deal.
Opt-out systems
Artists will, reportedly, have the right to opt out of having their work included in licensed training data.
However, some open-source generative systems (unlike Google's) have found this difficult to enforce, even when creators opt out. Recently, an artist who had opted out of Midjourney’s training data found that users had simply built a small training data set to continue mimicking his work. It is difficult to see how generative AI systems can work with legitimate IP unless they are centrally controlled. And even when they are, other systems could likely continue illegal imitations.
Central control raises questions too: while an artist of Drake’s clout might be able to withdraw their music (he was recently angered by recent tracks that copied his voice), it’s unlikely that a small act trying to make it will have the same freedom should this become a major money-maker for the label.
In context
The news follows an April request from UMG to streaming services, Apple Music and Spotify, to stop AI firms from ingesting the contents of their music libraries for “training” machines.
Across the entertainment industry, artists, workers, and creators are grappling with the profound and occasionally worrying ramifications of the technology, as the unions of screen writers and screen actors continue to strike in part over concerns about how studios use AI (among other grievances).
At Cannes Lions earlier this summer, Mattias Döpfner, CEO of the Germany-headquartered publisher Axel Springer took on the subject of IP in an AI age, declaring that “copyright as we know it is already gone.” In his talk, he posited the idea of “a quantitative model, like in the music industry, where [the AI platforms] have to prove which IP they’ve crawled, not that we have to prove that our IP has been crawled – that’s the main rule.”
Sourced from the FT, WARC
Email this content