EU Court Adviser: Chatbots Reusing News May Infringe; Training Is the Exception
In an advisory opinion on 2026-11-03, the EU court's Advocate General said Gemini's reuse of news content in certain circumstances may infringe, while training falls under an exception; the opinion is not binding, and was confirmed by two sources, Reuters and Dow Jones.
- 01THE OPINION — Advocate General Szpunar issues an advisory opinion on 2026-11-03, defining the circumstances in which chatbots reusing news content infringe.
- 02THE CONFIRMATION — Reuters and Dow Jones, republished by Morningstar, independently confirm the opinion's key points and the case background the same day.
- 03THE WATCHPOINT — The EU court's formal ruling and the Hungarian court's subsequent judgment.
WHAT HAPPENEDWhat Happened
Maciej Szpunar, Advocate General of the EU's highest court, issued an advisory opinion on 2026-11-03: chatbots driven by large language models that draw on news publishers' content to generate answers in certain circumstances may infringe news publishers' rights.
The case stems from a suit by Hungarian website operator Like Company against Google, centred on the Gemini chatbot's use of news content; the Hungarian court has sought guidance from the EU court in Luxembourg.
The opinion draws two key lines: first, reporting news facts themselves is not protected; what is protected is the specific form of expression of those facts. Second, obtaining content by bypassing a paywall or technical barriers, reusing the form of expression in an identifiable way, or leaving users with no need to read the original after reading the answer, may all fall into infringement territory.
The other side of the same opinion: using copyrighted material to train a large language model is a free use under the text and data mining exception; according to the court, there is no evidence that Like Company's articles were used to train the Gemini model.
WHY IT MATTERSTraining and Answering Are Split in Two
The most noteworthy thing about this opinion is not the conclusion itself, but that it splits training from answering. Training gets a pass, answering gets tightened, which means the compliance pressure on chatbot vendors is not in how models are trained, but in how answers are given.
For the news industry, the real variable is the ruling that follows: if the court adopts this standard, bypassing paywalls and identifiably reusing forms of expression become the starting point for licensing negotiations; if it does not, publishers must go back to fighting case by case in national courts.
SIGNAL JUDGMENT — ANALYSISThe Signal Is on the Answering Side, Not the Training Side
Tokenta Signal's judgment: if the court adopts this opinion, the compliance battleground for chatbots will shift from training data to answer output, with paywalls and reuse of expression the first two red lines to be tested. The next signal is the EU court's formal ruling and the Hungarian court's subsequent judgment.
WHAT REMAINS UNCERTAINWhat Remains Uncertain
EVIDENCEEvidence
Confirmed by two sources, the original Reuters report and independent Dow Jones reporting republished by Morningstar; the original link was opened and verified.
PRIMARY SOURCESPrimary Sources
- Reuters report
https://www.reuters.com/world/googles-use-press-by-ai-chatbot-may-infringe-copyright-rules-eu-court-adviser-2026-11-03/ - Dow Jones report republished by Morningstar
https://www.morningstar.com/news/dow-jones/202611035398/eu-court-adviser-warns-ai-chatbots-could-risk-infringing-copyright-law
RELATEDRelated Intelligence
- Update: Apple Intelligence Beta Goes Live in China, With Qwen Inside Siri
- Nvidia Reportedly Tells Cloud Partners to Cap Chinese Customers at 49% of New Contracts
- China Accelerates Full-Lifecycle AI Legislation: DeepSeek, Moonshot AI and Alibaba in Scope
- OpenAI and Anthropic set up election war rooms for the first time: the midterm election day duty test