A EU court adviser, Advocate General Maciej Szpunar, has raised the future possibility of copyright liability for chatbot in a lawsuit over allegedly stealing journalism and industry hard work
Szpunar’s advice is that publishers’ rights may be infringed when AI-generated answers reproduce protected parts of people’s work and news articles.
The findings could help shape legal boundaries for AI companies in a growing wave of copyright disputes over the use of protected material to develop and operate generative AI systems.
“When an LLM-driven chatbot retrieves a publication from a press publisher’s website, reuses that publication to generate an answer, and provides that answer to the user, the press publisher’s right may be infringed in certain circumstances,” Szpunar said.
The opinion concerns the case in Like Company v. Ireland Google, which has now become the first generative AI case to reach the Court of Justice of the European Union.
Like Company, a Hungarian operator of news web portals, alleges that Gemini repeatedly scrapped and reproduced its articles and made them available to the public through its AI chatbot’s responses between June 13, 2023, and May 31, 2024
A Budapest court paused the proceedings and referred four questions on EU copyright law to Luxembourg.
The case swept through a wider confrontation between rights holders and AI companies where, by late 2026 alone, more than 300 AI-related lawsuits had been documented, around 168 of them copyright claims, brought by authors, publishers, artists and music companies over the use of their protected works to train AI systems.
Realistically, the fear is real, as it reflects a wider commercial concern: AI chatbots can retrieve information from news websites, in some cases allegedly after bypassing paywalls, and deliver answers directly to users, allowing them to obtain information without visiting the original publication.
Publishers fear is that loss of visits can reduce referral traffic and advertising opportunities and undermine the financial value of their journalism.
Advocate Szpunar’s opinion only addresses copyright-related rights, not traffic losses, but hiding facts it combusts the same underlying question of who controls and commercially benefits from journalism when it is reused by generative AI.
But, does not treat the news itself as protected. It protects “only the specific form of expression” of a press publication, not the facts it reports, and is intended to safeguard the investment in producing it rather than to give publishers a monopoly over the news, he said.
For Gemini, that could mean liability where its AI chatbot answers reproduce a recognisable portion so closely that readers no longer need the original, especially where the material was obtained by bypassing a paywall or other technical barrier, or where such reproduction appears systematically in answers to related queries.
News publishers have been among the most consistent participants in sueing AI companies. Three years ago, The New York Times sued OpenAI, alleging millions of its articles were used to train models and that ChatGPT sometimes reproduces them nearly verbatim
Recently this year, October, USA Today Co. and its affiliates also sued OpenAI for more than $250 million, where it says hundreds of thousands of articles from 19 publications were copied for training.
Canadian publishers have also sued OpenAI and Anthropic, while other papers have challenged the scraping and distribution of news content through AI services.
Some publishers, meanwhile, like Reuters, the Associated Press, the Financial Times and Axel Springer have signed licensing deals with worthy millions or dollars, to let AI companies model scrap and use their contest for AI training
On the separate issue of training, Szpunar advised the court to declare the questions inadmissible in this case, saying there was no evidence Like Company’s articles were used to train Gemini and that training is not an “online use” of press publications
The use of protected material to train large language models is “freely permitted under the text and data mining exception,” he added
The opinion is advisory and not binding. The Court of Justice will now decide whether to adopt its reasoning, and the Hungarian court will then resolve the underlying dispute.
