Google and Universal Music Group are in talks to develop an AI-powered tool for users to create “deepfake” music using other artists’ existing work, the Financial Times reports.

The deal would reportedly pay copyright owners for the use of their likeness, and they would have the option to grant Google and UMG permission to license their work for the purpose of AI-driven replication.

The news arrives following Google’s release of MusicLM, a powerful generative AI tool allowing users to create music by entering prompts. Considering the company would need to license UMG’s music for creators to reproduce its vocals, lyrics and sounds, MusicLM is most likely a linchpin of the deal.

The deal is somewhat of an about-face for UMG, which controls roughly one-third of today’s global music market. Back in April, the company urged streaming platforms to block AI platforms from being able to leverage their music catalog when training their large language models.

If the two sides ultimately come to an agreement, it could lead to a watershed moment for the controversial nexus of generative AI and music. The scale of the framework would dwarf that of TuneCore, who recently partnered with Grimes to launch a pilot program enabling artists to utilize her deepfaked vocals “without penalty” and distribute the music to streaming platforms for a royalty split.

Meanwhile, Warner Music Group’s CEO, Robert Kyncl, reportedly condemned the proliferation of deepfake technology in the company’s Q3 conference earnings call this week. He said he believes artists should be the ones deciding whether or not their likenesses may be used to create music.

“There’s nothing more precious to an artist than their voice,” Kyncl said, “and protecting their voice is protecting their livelihood and protecting their persona.”