Google chatbot ‘may be infringing press rights’
An adviser to the EU’s top court has said that AI chatbots may be infringing the rights of press publishers when re-using their online publications to generate answers to queries from users.
Advocate General Maciej Szpunar was giving his opinion on questions referred to the EU’s Court of Justice (CJEU) by a Hungarian court.
The case is the first involving generative AI to reach the court.
Like Company, a Hungarian operator of web portals, had brought proceedings in a Budapest court claiming protection over the articles published on its portals under the copyright-related right granted by EU law to ‘press publishers’ over the ‘online use’ of their ‘press publications’.
The operator argued that its articles were used by Google to train the tech giant's Gemini LLM, and that the Gemini chatbot repeatedly reproduced the articles over the period from 13 June 2023 to 31 May 2024.
‘Specific form of expression’
Advocate General Szpunar found that when an LLM-driven chatbot retrieved a publication from a publisher’s website, re-used that publication to generate an answer, and provided that answer to the user, the publisher’s right might be infringed in certain circumstances.
He pointed out that the publisher’s right did not protect the facts reported in such articles, but only “the specific form of expression” of those publications.
In the advocate’s view, whether such use of press publications by an LLM-driven chatbot fell under the publisher’s right of reproduction depended on two factors:
- Whether the chatbot accessed the press publication at issue unlawfully – in particular by circumventing effective technical measures such as a paywall or by web-scraping, and
- Whether the answer provided to the user merely reported facts contained in the original publication or re-used the form of expression of that publication “in a recognisable manner, to such an extent that the user, by reading that answer, is entirely dispensed from accessing the original publication”.
The advocate said that the publisher’s rights were, as a rule, infringed when the publication’s form of expression was “systematically recognisable” in the answers provided by the chatbot.
Training questions
The Hungarian court also referred questions about the use of publishers’ articles for training LLMs, but the advocate declared these inadmissible, as there was no evidence that Like Company’s articles had been used to train the Gemini LLMs.
He suggested, however, that the use of material protected under copyright and the related right to train an LLM was, as a rule, freely permitted under EU law’s text and data-mining exception.
The advocate’s opinion is not binding on the court, which will give its final decision at a later date.