Artificial Intelligence in the Age of Copyright: The EUIPO Report

In recent years, rapid developments in artificial intelligence technologies have led to profound transformations in content creation. In particular, Generative AI (GenAI) systems are now capable of producing a wide range of content — from works of art and written text to music and visual media — with little to no human intervention. However, this transformation brings with it significant legal and ethical debates. The use of copyrighted content as training data by GenAI, and the legal status of the newly generated content, continue to be key issues under discussion worldwide.

EUIPO Report on the Development of Generative AI from a Copyright PerspectiveTo better understand these complex issues and assess how they can be addressed within a regulatory framework, the European Union Intellectual Property Office (EUIPO) has published a comprehensive study. Titled “GenAI from a Copyright Perspective,” the report sheds light on the technical functioning of AI-generated content and provides a detailed analysis of how existing and emerging EU copyright laws relate to GenAI.The primary aim of the study is to offer technical insights, legal analysis, and practical  recommendations for policymakers, rights holders, developers, and intermediaries.

Below is a summary of the key takeaways from the report.

The Development of Generative Artificial Intelligence from a Copyright PerspectiveIn the European Union, copyright plays a critical role in supporting cultural diversity and economic strength. Throughout the digital transformation, the sustainability of the creative sectors has been largely ensured through copyright protections. However, Generative Artificial Intelligence (GenAI) technologies are increasingly blurring the lines between human and machine-generated content, presenting new challenges to the existing copyright framework.

Key debates revolve around how copyright-protected content is used in the training of GenAI, how rights holders can assert and protect their rights in this process, and how AI-generated outputs should be identified and licensed. Legal and technical developments in this area demand balanced solutions that both safeguard creators' rights and foster innovation.

The EU aims to lead in digital transformation and plans to establish a Copyright Knowledge Centre by 2025, offering guidance and collaborative platforms for rights holders. In the near future, the question of whether AI-generated content itself should qualify for copyright protection is also likely to come into focus. The European Union Intellectual Property Office (EUIPO) intends to support this evolving landscape by providing technical solutions and mediation services to help uphold copyright protections.

An Overview of the Development of Generative AIIn recent years, artificial intelligence technologies have advanced rapidly. In particular, Large Language Models (LLMs) and Generative AI (GenAI) systems have enabled the creation of synthetic content, including text, code, images, audio, and video. These developments call on lawmakers and policymakers to revisit existing legal frameworks and ensure a balanced approach between intellectual property rights and technological innovation.

Technical Background

GenAI systems develop their algorithms by utilizing large volumes of training data. The training process typically involves stages such as data collection, preprocessing, pretraining of models, fine-tuning, and reinforcement learning. While initially relying on limited and labeled datasets, today's models are often trained using vast datasets gathered through web scraping techniques. The type of content generated varies depending on the model and underlying technologies used. For real-time content updates, Retrieval-Augmented Generation (RAG) technologies may be employed.

Regulatory and Legal Background

In the European Union, two major legal frameworks govern GenAI: the Copyright in the Digital Single Market Directive (CDSM) and the Artificial Intelligence Act (AI Act). The CDSM provides copyright exceptions for text and data mining (TDM) for scientific research and other purposes, while allowing rights holders to “opt out” and reserve their rights by prohibiting the use of their content. The AI Act mandates that providers of general-purpose AI systems disclose detailed information about training data and ensure that AI-generated content is machine-detectable.

Due to the growing use of GenAI, legal disputes between rights holders and AI developers are emerging globally, particularly in the United States and Europe. In the EU, lawsuits have already been initiated in countries like Germany and France.

Economic Background

Licensing agreements are being negotiated between rights holders and GenAI developers. Developers are generally motivated to obtain licenses due to data scarcity and the need for high-quality content. The press and academic publishing sectors hold a competitive advantage in licensing opportunities through the application of RAG. For smaller rights holders, the emergence of content aggregation intermediaries expands licensing possibilities. The growth of the licensing market depends on the establishment of standard pricing models and payment terms. Direct licensing may offer new revenue streams, but rights holders must actively exercise their TDM rights to benefit.

Input for Generative AI

Due to widespread web scraping, rights holders increasingly employ opt-out mechanisms to prevent the unauthorized use of their content. The Robots Exclusion Protocol (REP) is a commonly used method but is often viewed as inadequate. A combination of legal and technical measures—including unilateral declarations, licensing conditions, and technical protocols—is being utilized. Each method has its own advantages and limitations, and stakeholders are calling for standardized yet flexible solutions. Public authorities are expected to play a role in managing TDM reservations and raising awareness.

Generative AI Output

The Artificial Intelligence Act mandates transparency for content generated by GenAI systems. This study explores methods developed to detect fake content and trace its origin. To reduce the risk of copyright infringement, GenAI providers are developing tools for content comparison, filtering, and model updating. Some providers also offer legal warranties to limit liability.

Public authorities can guide AI developers and policymakers on detecting synthetic content, mitigating infringement risks, and promoting ethical use. They can also raise public awareness and create platforms for technical collaboration.

Findings

As GenAI models evolve, access to high-quality and up-to-date content becomes increasingly important. While many developers still use content without permission from rights holders, a direct licensing market is gradually emerging. Effective opt-out solutions are needed to support this market, though each current method has its own pros and cons.

Transparency is essential for identifying content sources, clarifying permissible uses, and distinguishing AI-generated outputs. This facilitates copyright enforcement for both rights holders and GenAI developers.

This study highlights the technical and legal complexities GenAI introduces to creative and cultural sectors. The EUIPO Copyright Information Hub, to be launched by the end of 2025, offers an opportunity to address these complexities and guide rights holders.

Moreover, the study emphasizes the need for closer collaboration between AI developers and creative industries to comply with TDM opt-out requests. The Information Hub is expected to provide platforms for licensing and mediation, raise awareness of copyright risks, and support ongoing discussions on protecting AI-generated content.

You can access the full report at the following link: https://www.euipo.europa.eu/en/news/euipo-releases-study-on-generative-artificial-intelligence-and-copyright