EU court adviser says Google's Gemini may breach press copyright rules
A legal opinion from Europe's top court adds scrutiny to how artificial intelligence tools use news content from publishers' websites. The adviser says Google's Gemini may infringe press publishers' rights in some cases, while use of copyrighted material to train large language models remains permitted under an EU exception.
Highlights
- EU Advocate General Maciej Szpunar states Google's Gemini chatbot may infringe publisher copyright when reproducing protected press content in user answers.
- Szpunar asserts EU copyright law allows use of press publications for large language model training via the bloc's text and data mining exception.
- The legal opinion, issued in a Like Company v. Google case, suggests AI-generated answers face greater legal risk than model training, pending CJEU judges' final decision.
Legal opinion outlines limits on chatbot use
As reported by Reuters, Advocate General Maciej Szpunar says a press publisher's rights may be infringed in certain circumstances when an LLM-driven chatbot retrieves a publication from a publisher's website, reuses it to generate an answer and then provides that answer to a user.He says EU copyright rules do not protect the mere facts reported in press publications, but only the specific form of expression. At the same time, he says the use of copyrighted material to train large language models is freely permitted under the bloc's text and data mining exception.
The opinion comes in a case brought by Hungarian website operator Like Company against Google over the use of press publications by the company's Gemini chatbot.
Potential impact on publishers and AI platforms
The assessment points to a narrower legal risk for AI-generated answers than for model training, drawing a distinction between reproducing protected expression and processing content under existing data-mining rules.Opinions from advocates general are not binding, but the judges of the Court of Justice of the European Union often follow them. A final ruling could help clarify how EU copyright law applies to generative AI services that draw on publisher content.
In our earlier article, we covered ByteDance’s move to set up an Alignment & Preparedness team inside its Seed research unit and launch a global hiring drive focused on frontier-model safety. We noted that the new group is intended to strengthen testing, monitoring and risk preparedness as AI systems become more autonomous, reflecting growing regulatory and cross-border scrutiny of how advanced models are developed and deployed.