Sony Music sues Udio, again.

On Monday, July 20, 2026, Sony Music Entertainment filed a new copyright infringement lawsuit against Udio Inc. in the U.S. District Court for the Southern District of New York (SDNY). The complaint specifically lists 30,117 tracks Udio allegedly used without a license to train its models. Reported artists whose recordings are named include Beyoncé, Harry Styles, and Elvis Presley (Innovatopia).

This case extends a lineage that began in June 2024, when Universal Music Group (UMG), Sony Music, and Warner Music Group (WMG) — coordinated by the RIAA — sued both Udio and Suno. That original complaint named just 333 tracks. During subsequent discovery, Sony said sound-signature matching turned up a far larger number of its catalog tracks in Udio's training data, and it moved to add those to the existing case. But on June 29, 2026, Judge Alvin K. Hellerstein denied that motion, effectively requiring Sony to bring the claims as a separate action. This new complaint is the direct result of that ruling (Music Business Worldwide).

In 2024, all three major labels — UMG, Sony Music, and WMG — moved in lockstep against Udio and Suno. But in October 2025, UMG broke ranks first, settling and pivoting toward a joint, fully licensed AI music platform. WMG followed. That leaves Sony as the only major label still fighting Udio in court over unlicensed training data.

Seen in that light, this re-filing isn't just about a bigger number. It is the one remaining case that could settle, by judgment rather than settlement, the central question hanging over generative AI and copyright: is unlicensed training-data use infringement, or a permissible technical step toward producing a new, non-infringing work? UMG and WMG's settlements pointed the industry toward licensing as the pragmatic outcome; Sony's lawsuit is the venue where that outcome gets legally tested.

In the near term, watch Udio's answer to the new complaint. In its April 2026 response to the original suit, Udio admitted to training on copyrighted songs uploaded to YouTube, but argued that this amounted to fair use because it was "a back-end technological process, invisible to the public, in the service of creating an ultimately non-infringing new product." Expect a similar fair-use defense here.

Three recommended actions for anyone using AI music generation commercially: (1) re-check the scope of your Udio usage and any indemnification terms in your contract; (2) track this case's trajectory — settlement versus a fair-use ruling — on a quarterly cadence; (3) factor in the possibility that Sony eventually follows UMG and WMG into a licensing deal, and evaluate licensed alternatives in parallel as a hedge.

This isn't a story to read with unqualified optimism for either side. First, Udio's fair-use argument has real legal grounding, and U.S. case law on training-data use for generative AI is still unsettled — Sony's claims are not guaranteed to prevail. Second, the precision of Sony's identification method — sound-signature matching, plus the yt-dlp allegation — will itself be contested through further discovery and argument. Third, with UMG and WMG already settled, the industry's center of gravity is shifting toward licensing over litigation, so the outcome of Sony v. Udio won't necessarily generalize to other AI music generators such as Suno or to industry practice at large. A ruling is likely months if not longer away, so plan around uncertainty rather than a fixed outcome.