AI Training, Chatbot Outputs and Copyright: A Pending CJEU Case with Major Implications for Businesses

Introduction

A preliminary ruling request currently pending before the Court of Justice of the European Union (CJEU), initiated by a Hungarian court, raises fundamental questions regarding the use of copyrighted content by artificial intelligence systems. The case, brought by a Hungarian publisher against Google, may significantly reshape the legal framework governing AI use in business and marketing contexts across the EU.

Factual Background

The claimant, a Hungarian news publisher, alleges that its articles were used without consent and without remuneration:
– for training Google’s large language model (Gemini), and
– within the outputs of the Gemini chatbot when responding to user prompts.

The national court referred multiple questions to the CJEU, essentially asking whether such uses infringe the publisher’s rights under EU copyright law, in particular under the DSM Directive.

Two Distinct Uses: Training vs Output

The case distinguishes between two technically and legally different phases:

Training phase
During training, content is transformed into tokens and statistical relationships, forming model parameters. Google argues that no direct reproduction occurs in a legally relevant sense.
Output phase
The chatbot generates responses probabilistically. However, it was acknowledged during the hearing that outputs may, in certain cases, reproduce or closely resemble original protected texts.

This distinction is central, yet the European Commission emphasized that legal qualification should not depend on technical abstraction.

Key Legal Issues

Reproduction and communication to the public
If AI outputs reproduce protected content, even partially, this may constitute copyright infringement.
Text and data mining (TDM) exception
A critical question is whether the TDM exception applies to commercial AI systems.
– A restrictive interpretation would require licensing
– A broader interpretation would favour AI developers
Burden of proof
Given the opacity of AI systems, the issue arises whether service providers must demonstrate that no infringing use occurred.
Territorial scope
Even if training occurs outside the EU, EU law may apply where outputs target EU users.
Publisher’s rights vs access to information
The case also highlights the tension between:
– protecting content creators and publishers, and
– maintaining open access to information and innovation

Competition and Platform Dependency

An additional concern is whether publishers are indirectly forced to allow AI usage in order to maintain visibility in search engines. Such dependency could raise issues under the Digital Markets Act (DMA) and EU competition law.

Economic Impact

One of the most significant risks identified is the so-called “traffic substitution effect”:
If users obtain the substance of articles directly from AI systems, they may no longer visit the original websites, undermining publishers’ business models.

This could have systemic consequences for the sustainability of content creation.

Regulatory Context

The outcome of this case will interact with broader EU regulation, including:
– the AI Act
– the Digital Services Act (DSA)
– and the Digital Markets Act (DMA)

Together, these frameworks indicate a clear trend:
AI is becoming a regulated economic activity, not merely a technological tool.

Practical Implications for Businesses

Companies using AI in marketing and operations should consider:

– implementing AI governance frameworks
– conducting copyright and compliance audits
– reviewing AI provider contracts
– introducing human oversight over AI-generated content
– ensuring transparency in AI-assisted communications

Conclusion

The pending CJEU decision may establish a landmark precedent regarding the relationship between AI and intellectual property law in the EU.

Regardless of the outcome, one principle is already clear:

The use of AI does not eliminate legal responsibility.

Businesses leveraging AI must proactively address legal risks or face increasing exposure in a rapidly evolving regulatory environment.

Dr. Katona Géza, LL.M. ügyvéd (Rechtsanwalt / attorney at law)

___________________________________