On 10 March 2026, the Court of Justice of the European Union held its first-ever oral hearing on the question of whether training large language models on European-authored content constitutes copyright infringement. The case — involving a coalition of European publishers and creative rights holders against Google’s generative AI operations in Europe — is potentially the most consequential AI legal proceeding in European history. Here is what happened, what is at stake, and what GPAI providers must do to prepare for the ruling.
- The CJEU is hearing arguments on whether LLM training on European content is covered by the Text and Data Mining (TDM) exception in the DSM Directive — or whether it constitutes infringement requiring licensing.
- A ruling against the TDM exception for commercial LLM training would create an obligation to license training data from European rights holders — potentially making current LLM training methodologies on European content illegal retroactively.
- The EU AI Act’s Article 53 already requires GPAI providers to document their copyright compliance approach. This case will determine whether existing approaches are legally sound.
- A ruling is not expected until late 2026 at earliest. But the mere existence of the case changes your GPAI compliance posture now.
1. The Case: What Is Being Argued
The case before the CJEU is a reference from a German appellate court, which referred a set of questions to the CJEU following a dispute between a coalition of European news publishers, book publishers, and creative rights organisations (acting collectively through the STM publishers’ coalition and national collective management organisations) and Google LLC’s European operations.
The core allegation is that Google scraped hundreds of millions of European-authored articles, books, scientific papers, and other copyrighted works from European websites and databases to train its generative AI models — including the models underlying Google Search’s AI Overview, Google Bard/Gemini, and NotebookLM — without licensing this content from the European rights holders who own it.
Google’s defence is that this training falls within the Text and Data Mining (TDM) exception in Articles 3 and 4 of the DSM Directive (2019/790) — which allows reproduction of lawfully accessed content for TDM purposes, including for commercial purposes (with an opt-out mechanism for rights holders).
The publishers argue that generative AI training is not “text and data mining” within the meaning of the DSM Directive — it is a new form of exploitation that creates competing works and should require licensing.
2. The Legal Framework: TDM Exceptions Under the DSM Directive
The Directive on Copyright in the Digital Single Market (EU 2019/790) introduced two TDM exceptions:
The DSM Directive defines TDM as “any automated analytical technique aimed at analysing text and data in digital form in order to generate information which includes but is not limited to patterns, trends and correlations.” The publishers argue that training a generative AI to reproduce, synthesise, and recombine creative content is not merely “generating information” about patterns — it is creating a derivative product from the copyrighted works themselves, taking them far outside the definition of TDM.
3. What Was Argued on 10 March
The 10 March oral hearing before the CJEU’s Grand Chamber (19 judges — indicating the case’s constitutional significance) lasted approximately six hours. Neither party’s arguments are fully public, but based on reporting from observers present and statements from legal representatives, the key positions were:
The Advocate General is expected to deliver a non-binding opinion before the full Court ruling. Advocate General opinions are influential — the CJEU follows them in approximately 80% of cases. The opinion is expected in September 2026, with a full ruling likely in Q1 2027.
4. Possible Outcomes and Their Consequences
5. What This Means for Your GPAI Compliance Right Now
You do not need to wait for the CJEU ruling to take action. The EU AI Act’s Article 53(1)(c) already requires GPAI model providers to implement and make publicly available a policy to comply with Union copyright law. The CJEU case determines what “compliance” requires — but the obligation to have a policy and to document your approach is immediate. See our EU AI Act Summary 2026 for the full GPAI obligations framework.
6. Other Key Legal Cases to Watch
| Case | Jurisdiction | Issue | Status |
|---|---|---|---|
| News publishers v. Google (CJEU referral) | CJEU / Germany | LLM training and TDM exception | Hearing completed March 2026 |
| Authors Guild v. OpenAI | US (S.D.N.Y.) | Book training, fair use | Discovery phase, 2026 |
| GEMA v. Suno/Udio | Germany | Music training data | Pending first instance |
| SPAIn v. Meta (AI training) | Spain | Instagram image training data | Under AEPD investigation |
