Isbell's class action against Suno runs on 17 counts of state identity law

Twenty-four artists appear in the filing as worked examples, the four named plaintiffs plus the roughly 20 further names counsel walks through ^[1][14][21], and against that the pleaded floor for the whole proposed class works out to about $208,000 apiece ^[22], which is what a jurisdictional threshold looks like when it is read as a valuation ^[6]. The complaint also alleges Suno trained on tens of millions of recordings ^[19]. The two figures answer different questions: one sets a jurisdictional floor for the suit, the other describes the scale of the alleged training data.

The remedy section is more informative than the figure. Damages set by state law, disgorgement of the profits Suno made from the alleged conduct, and punitive damages where state law permits ^[6] describe a claim assembled one jurisdiction at a time. Per the account in pivotnews.ai of Fast Company's reporting, the case rests on identity misappropriation rather than the copyright claims that have driven most of the music industry's fight with AI generators ^[15][16]. Every count in the 84 pages arises under state identity law; copyright never appears ^[2]. (The complaint opens and closes on the Borg from Star Trek ^[17], a register chosen for a jury and a press cycle rather than for a motion hearing.)

The counts then turn on a factual question that discovery can answer cheaply. Suno has said publicly that it does not use artist names as a category of training metadata and that it built detection filters blocking artist, song and album names as prompts ^[8]; the complaint calls that claim false and says the model accepts names and answers them ^[9], alleging the filter can be beaten by putting spaces between the letters of a name, with tutorials from Suno's own paid affiliate marketers teaching subscribers the workaround ^[10]. Typing "jason isbell" into the v5 model, the plaintiffs say, returned an Americana track called Paper Bell with fingerpicked acoustic guitar and a cover image of a paper bell set against a church, imitating Isbell's clear male vocals and country twang ^[11].

The available defence sits inside that description, because fingerpicked Americana and a church on the artwork is a genre with a stock image attached, and whether the output identifies a person or evokes a style is the question these counts have to survive ^[11]. The second defence is procedural: damages measured by each state's own statute ^[6] is the standard ground on which a nationwide class fails certification. The damages arithmetic matters less than the injunction: an order barring Suno from capturing, storing or commercially exploiting musicians' voices, names and identifying attributes without consent ^[7] is a constraint on what the model may hold, not a licence fee it can budget for. If the logs show the name filter works as Suno described it and the cited outputs were engineered around it by users ^[8][10], the identity theory shrinks to a case about prompt hygiene and the injunction never issues.