EU Court Advisor: Google's Gemini May Infringe Copyright by Using News Content
nashnova research
The EU Court of Justice's top legal adviser ruled that Google Gemini may infringe publishers' copyright when it pulls news content to generate answers — but using the same content to train the AI model does not. Where that line lands will shape how every AI platform handles news.
What exactly did the adviser say?
Advocate General Maciej Szpunar issued an opinion on November 3: when Gemini retrieves content from news websites, generates an answer, and presents it to users, publishers' rights may be infringed in certain circumstances.
He stressed that copyright protects not the "bare facts" in a news report but the specific form of expression. This means → an AI restating a story's facts in different words may be fine; copying the original wording and structure could cross the line.
In plain terms = anyone can report the facts, but the article you wrote is copyrighted.
Why does training on news content not count as infringement?
Szpunar also found that using copyrighted material to train a large language model falls under the EU's "text and data mining exception" and qualifies as "free use" — not infringement.
This means → the adviser drew a clear line: training is permitted; generating answers may not be.
This reflects the EU's core stance on AI copyright: a machine "learning" content and a machine "reproducing" content are two entirely different acts.
Can this opinion take effect directly?
The case was brought by Hungarian website operator Like Company against Google and is currently before the EU Court of Justice.
The Advocate General's opinion is not legally binding, but EU court judges typically follow these opinions when issuing their rulings.
In plain terms = this is not yet a judgment, but it will most likely set the direction for one.
What does this mean for the AI industry?
The opinion's core distinction — the copyright boundary between AI-generated answers vs. AI training — would become an EU-wide precedent if adopted by the court.
This means → every AI platform that generates answers from news content (not just Google) may need to reassess how it handles publisher material, or negotiate licensing deals.
For news publishers, this is a critical legal weapon: it marks the first time an EU-level opinion has confirmed that the answer-generation stage carries copyright risk.
市场有风险，内容仅供研究参考，不构成投资建议。