Europe's AI copyright legal landscape looks very different from the US

[Merchant89]

A legal analysis from Dutch law firm Stibbe warns that technology companies training AI in Europe face a structurally different legal environment than the US, since fair use, the flexible doctrine that's central to ongoing American AI copyright cases, simply doesn't exist under Dutch or EU law. Unlike the US, where Anthropic became the first major AI company to settle a copyright class action, agreeing to pay authors 1.5 billion dollars in 2025, EU copyright exceptions are exhaustively listed and narrowly interpreted, meaning if a specific use doesn't fit one of those defined exceptions, it's infringement, full stop, with no judicial discretion to invent new exceptions

The EU's text and data mining exceptions allow scraping for AI training only if rights holders haven't opted out, typically through a machine readable robots.txt file, and a German court has already ruled against OpenAI in a case brought by music rights group GEMA, finding that when a model memorizes protected works and can reproduce them in outputs, that goes beyond what the exception permits. A major EU top court ruling on text and data mining in generative AI specifically is expected in 2027, following an Advocate General opinion due this month

The analysis warns that the Netherlands' WAMCA mass litigation law, which has already produced more than 100 collective claims worth tens of billions of euros including privacy claims against TikTok, gives rights organizations a ready made legal infrastructure to bring large scale AI copyright claims, and that developing case law on EU-wide jurisdiction means such a claim may be a question of timing rather than probability. Curious what people think this means practically for AI companies operating globally, does Europe's stricter framework meaningfully change how these companies actually train their models, or mostly just increase legal costs without changing underlying practices

AI Copyright Claims in Europe: Key Considerationswww.stibbe.com

[Archie91]

WAMCA already having produced over 100 mass claims worth tens of billions shows the legal infrastructure for a massive AI copyright case genuinely already exists in the Netherlands, this isn't a hypothetical legal mechanism

[BlueFalcon]

Curious whether this actually changes how any major AI lab trains its models going forward, or whether companies just treat European legal exposure as an acceptable cost of doing business globally

[RomanReigns02]

The German ruling against OpenAI specifically drawing a line at models that can reproduce memorized protected works feels like a genuinely important and fairly narrow technical distinction other courts will likely adopt

[Benzema]

No fair use equivalent at all in EU law is honestly the detail that should worry AI companies most, an American court ruling favorably on fair use provides zero legal protection once you're actually operating in Europe

[Gareth_11]

Anthropic's 1.5 billion dollar settlement being cited as a comparison point shows how much money is genuinely at stake industry wide here, that's a real number other courts and rights holders will keep referencing

[WorldClassHart13]

Robots.txt functioning as the primary opt-out mechanism for training data feels like a fairly blunt tool for something with this much legal weight now riding on it