Text generation: AI training with protected works becomes a case for the ECJ

The European Court of Justice (ECJ) has to deal with some of the most pressing issues surrounding artificial intelligence (AI) and copyright at the moment. A Hungarian court has referred questions to the ECJ regarding the case "Like Company vs.

Gemini and copyright law

The case concerns a lawsuit brought by the Hungarian publishing house Like, which operates the ad-financed news site balatonkornyeke.hu, among others. The publisher accuses Google that its AI chatbot Gemini has reproduced a copyrighted article about a planned dolphin aquarium on Lake Balaton as a more or less verbatim summary.

The Budapest District Court (Környéki Törvényszék) is now asking the ECJ to clarify several questions. For example, it wants to know whether the display of content in a chatbot that is identical to the protected content on a publisher's website is an act of reproduction and making available to the public.

Should the Luxembourg judges answer in the affirmative, this would likely constitute an infringement of copyright and the special ancillary copyright for press publishers on the internet. In this case, the Budapest court also wants to know whether it plays a role that chatbots such as Gemini only predict the next word in their responses on the basis of observed patterns.

The Budapest judges are also more fundamentally concerned with the question of whether the training of a system for generative AI falls under the right of reproduction. In the EU, legislators have established exceptions to the exclusive exploitation right for text and data mining in the latest copyright amendment. The Hungarian court wants to know whether this copyright barrier is applicable in the case in question.

Only reference to facts by Gemini?

The publisher accuses Google of continuous legal infringements because the search engine giant repeatedly uses protected texts without consent. The scope of use exceeds the use of "individual words or very short excerpts" covered by ancillary copyright. Consent to the display of content in search engines does not apply to any further distribution. Google had infringed the reproduction right in its training.

Google countered that it was acting within the scope of the copyright restrictions granted. The response generated by Gemini does not constitute a case of making available to the public or reproduction. Gemini merely refers to some of the facts contained in the protected content.

The EU Intellectual Property Office (EUIPO), however, sees evidence that most AI systems "obtain and use content available online without the prior authorization of copyright holders". The experts at the specialist blog IPKat also assume that the operation of a chatbot affects the rights of reproduction and making available to the public.

A ruling by the ECJ is unlikely before 2027.

(mma)