The Frankfurt Higher Regional Court has upheld a lyricist's injunction against an "it was AI" defence

Key takeaways

• The court held that where a raw lyric was the claimant's own creation and its essential creative elements remain in the published version, the question of later AI use for the published version does not matter (the court's words: "kommt es auf die Frage des späteren Einsatzes eines KI-Systems für die veröffentlichte Fassung nicht an"; English paraphrase ours).
• The claimant made her authorship credible with sworn statements, which the court found convincing in themselves.
• The court was not persuaded by the indicators of AI generation in the defendant's private expert report.
• The press release says: "Die im Eilverfahren ergangene Entscheidung ist nicht anfechtbar."

What did the Frankfurt court decide?

The Oberlandesgericht Frankfurt am Main, 11th Civil Senate, confirmed an injunction for the claimant as author of a song lyric. The court's press office (press release 52/2026, 20 October 2026) says the claimant is entitled to have the defendant stop using the lyric without authorisation. The judgment is dated 1 October 2026, file number 11 U 2/26. The press release says the first-instance Landgericht Frankfurt am Main had granted the application and that the appeal "had essentially no success".

The press release states the central holding in one sentence: if it can be shown and made credible that a lyric in its raw version arose as the author's own creative work, the later use of an AI system for the published version does not matter, provided essential creative elements of the raw version remain in the published version.

What were the facts, and what did each side argue?

According to the press release, the claimant's partner released a song on the platform DistroKid. The claimant says she wrote the underlying original lyric herself. Her partner used the AI system the press release calls "Suno-AI" in producing the music. Three months later the defendant released two versions of a song on Spotify and elsewhere. The press release says the artist behind that song promoted it by telling listeners to come in "before the song gets blocked", and said she was going on tour with a song that did not belong to her.

The claimant applied for interim relief, asking among other things that the defendant stop releasing the song because her lyric had been used without permission. The defendant argued, among other points, that the lyric had been created by AI and not by the claimant. This is a dispute about authorship of the words.

How did the court weigh the AI evidence?

The press release says the claimant made her account of creating the raw version credible through "in itself convincing" sworn statements. The defendant relied on a private expert report listing indicators of an AI-generated lyric. The court was not convinced by them. It gave two examples. The report treated it as unusual that the chorus is performed in full only once and that everything seems "exaggerated"; the court said it was left open why that points to AI generation, and, absent explanation, it tended to point against digital generation given that AI software usually works on probabilities and pattern recognition. The report also said the sentences in combination sometimes made "little sense"; the court found a clear narrative thread, which the expert himself had recognised by reference to the underlying "Cinderella plot".

Why did later AI use not decide the case?

The press release says the court left open whether and how far the first text version was later edited, in whole or in part, by the software it calls "Sono-AI" while text and music were adapted. The court found that the decisive creative elements of the raw version were still present in the version finally published. Because the defendant's published lyric contained recognisable, essential creative parts of the claimant's work, the court held that the defendant had unlawfully interfered with the usage rights vested in her as author.

The press release also carries an explanatory note. It says only personal intellectual creations of a human being are protected under section 2(2) of the German Copyright Act (UrhG), that under the creator principle in section 7 UrhG only a human can be an author, and that a text generated exclusively by an AI system therefore has no copyright protection. It adds that there is so far little case law on whether and when elaborate prompting makes an AI system's result a human work. We treat that note as the press office explaining the law, since the court did not need to decide the point on these facts.

What should music, publishing and creator teams do now?

The case shows what carried the day: a human raw version, credible sworn statements, and a defendant whose expert evidence did not explain why the features it pointed to indicated AI. Teams cannot assume the same result in a dispute with different evidence.

• Keep dated drafts, notes and files for the human raw version of lyrics, scripts and other text, separate from later AI-assisted versions.
• Record which AI tools were used at which stage, including tools used only for music or production.
• If you receive a claim, find out whether the opponent says the work is AI-generated, and ask what evidence supports that. A bare list of stylistic indicators was not enough here.
• Read the full judgment when it is published before relying on it in a filing; this article rests on the court's press release.

What are the key dates and decisions?

What we did not verify

What we opened: the Hesse judiciary press release 52/2026 of the Oberlandesgericht Frankfurt am Main, dated 20 October 2026, on 2026-10-25.

What we did not open: the full judgment (the press release says it will be available shortly on the Hesse legal database), the first-instance judgment of the Landgericht Frankfurt am Main, the defendant's private expert report, and the parties' filings. Every fact here is as reported in the press release, which names the AI tool both as "Suno-AI" and "Sono-AI".

What we refuse to claim: that this is a general rule on AI-generated works, that AI-assisted text is protected in every case, who the parties are, or any damages, costs or further proceedings.

Related briefings

• AI Regulation News hub
• Europe AI regulation tracker
• Attorneys and AI regulation
• Munich court, GEMA and Suno on memorisation
• Munich court, GEMA and OpenAI on AI training and copyright
• Hamburg appeal court, LAION and the text and data mining opt-out
• US Copyright Office on AI training and human authorship