What’s solid here, and corroborated across JUVE Patent, MusicTech, MusicWeek and DJ Mag’s reporting on the court file: Munich’s 42nd Civil Chamber ruled on July 31, 2025 that Suno directly infringed copyright by training its model on six specific GEMA-repertoire tracks — including Boney M’s “Daddy Cool” and “Rasputin,” Alphaville’s “Forever Young” and “Big in Japan,” Helene Fischer’s “Atemlos” and Lou Bega’s “Mambo No. 5.” GEMA filed the case on January 21, 2025, and the court found Suno used stream-ripping to pull the tracks off YouTube, circumventing a technical protection measure. Suno’s US-based training activity does not shield it in Europe, the court held — a first for any European court on that question. Suno is ordered to cease and desist, disclose revenue, and pay damages in an amount still undetermined; the judgment is not yet enforceable, and Suno told MusicWeek and DJ Mag it disagrees and is “evaluating all available options, including an appeal.”
Where the story gets softer is APRA AMCOS’s extrapolation from that German ruling to Australian catalogs. CEO Dean Ormston’s claim that members’ songs from Midnight Oil, Cold Chisel, Sia, Crowded House and Lorde turned up “uninvited and unpaid” inside Suno’s system is not backed, in anything reviewed here, by the kind of forensic showing GEMA presented in Munich — where expert testimony demonstrated melody, harmony and rhythm matches between Suno outputs and specific compositions. Ormston’s list reads as an assertion of harm, not a documented finding; nothing in the independent reporting confirms these artists’ works were used in Suno training or reproduced in outputs. That distinction matters for anyone assessing whether Australia is headed toward its own GEMA-style suit or whether this is advocacy framing ahead of one.
A German court decision doesn’t automatically become Australian law, and a trade body’s list of allegedly-copied artists isn’t the same as a court-tested forensic match.
Also worth flagging as a claim rather than a fact: Suno’s own defense, relayed to MusicWeek, that its models are trained “to create new songs, not reproduce existing ones.” The Munich court didn’t buy that characterization for the six songs at issue, finding the works were effectively stored and reproducible via memorization — a technical determination JUVE Patent’s account describes in some detail. That’s a live dispute Suno intends to appeal, and it’s the crux of the fair-use-adjacent argument Suno will need to win elsewhere, including in its pending US litigation with major labels.
What would validate APRA AMCOS’s broader claim: an Australian rightsholder action that produces the same kind of side-by-side acoustic matching GEMA supplied in Munich, or evidence from Suno’s disclosed revenue and training logs — now compelled in Germany — showing Australian repertoire was ingested. Absent that, the practical takeaway for data licensing teams is narrower than Ormston’s rhetoric suggests: a German court has established, on a specific evidentiary record, that AI training abroad doesn’t insulate a company operating servers in the EU, and that GEMA’s 100,000-plus members (and the two million rightsholders it says it represents globally, per MusicTech and MusicWeek) now have a template. Whether that template extends to Suno’s reported $5.4 billion valuation and its other pending suits, per DJ Mag, is the next thing to watch — along with GEMA’s damages figure, still undetermined, and Suno’s promised appeal.
“We already know what has been taken. Midnight Oil. Cold Chisel. Sia. Crowded House. Lorde. Our members found their own songs inside these systems, uninvited and unpaid. This is not just a German case. It's a warning to every AI developer, in every territory, including this one.”
