Suno Loses Most of Its Motion to Dismiss

Suno Asked to Dismiss Three Counts. It Won One, and the One It Won Was Abandoned Anyway.

Judge Saylor held that a copyright plaintiff need not produce the infringing output at the pleading stage — the point on which AI music cases have been turning.

In Justice v. Suno, Inc., No. 25-11739 in the District of Massachusetts, Judge F. Dennis Saylor IV ruled on 20 August that the AI music company's motion to dismiss is granted as to Count 4 and otherwise denied. The plaintiffs are independent recording artists and artist-owned entities holding copyrights in compositions and recordings. They allege Suno downloaded their songs from YouTube, copied them to train its model, and generates music substantially similar to protected expressive elements of those songs.

The holding that matters

Suno's central argument was that plaintiffs must point to a specific infringing output to state a claim. The court rejected it: "there is no categorical requirement that a plaintiff must produce the derivative work at the motion to dismiss stage." Saylor distinguished the cases Suno cited, noting that in those the courts had been unable to meaningfully assess anything beyond bare legal conclusions. The amended complaint asserts four counts — reproduction under §106(1), derivative works under §106(2), circumvention under the DMCA at §1201(a)(1), and the Tennessee consumer-protection claim.

What the common framing gets wrong

Reporting a "partial win for Suno" inverts the result. Suno moved against three counts and lost on two of them, including both copyright counts and the DMCA count — the substance of the case. The count it won was Count 4, which failed on a pleading defect rather than on the merits: plaintiffs pleaded the Tennessee Consumer Protection Act and then argued in opposition that they had really meant a Tennessee common-law unfair-competition claim, and the court declined to evaluate a theory that was not pleaded. Note also what an order like this is not. Denying a motion to dismiss decides only that the allegations, assumed true, could support relief. Nothing here holds that Suno infringed anything.

Where it leaves the AI music cases

Requiring a matching output at the pleading stage would have been close to fatal for training-data claims, since plaintiffs cannot run a model they do not control against every song they own. Saylor declining to impose that requirement keeps the case alive into discovery, which is where the training corpus becomes visible.

Why It Matters

Whether copyright plaintiffs must show an infringing output before discovery is the pivotal procedural question in AI training cases. A district court refusing to make it a threshold requirement moves several suits past their most dangerous gate.