Legal Harmonization of Intellectual Property Rights in Europe: Balancing National Sovereignty and Regional Integration

Have you got a question?

The protection of intellectual property (IP) rights in Europe has evolved into a cornerstone of the region’s knowledge-based economy. Despite substantial progress through institutions such as the European Patent Office (EPA) and the EU Intellectual Property Office (EUIPO), persistent national legal differences, coupled with the rapid emergence of technologies like artificial intelligence (AI), continue to complicate the European IP landscape.

I. Harmonization of Intellectual Property Rights in Europe

In Europe, the EPA and EUIPO have spearheaded efforts to harmonize IP registration processes. Similarly, the EU’s Digital Copyright Directive seeks to standardize the legal landscape for digital rights, but national exceptions for purposes such as education and parody highlight the limitations of harmonization efforts..

Trademarks and designs have achieved relatively greater consistency due to the EU Trademark (EUTM) and Community Design systems, which allow businesses to secure region-wide protections through a single application. However, even in these areas, enforcement remains a challenge, requiring businesses to navigate diverse legal frameworks across member states. While these harmonization efforts have improved consistency, the rise of AI-driven innovations exposes unresolved legal gaps.

1. Patenting AI Innovations: Challenges and Legal Gaps

The rise of AI has created new legal frontiers in patent law, exposing gaps in established frameworks that were designed with human inventors in mind. AI innovations frequently involve highly complex algorithms, autonomous processes, and machine learning models that challenge traditional notions of inventorship, technical disclosure, and patent eligibility.

1.1 Inventorship and Human-Centered Legal Models

Current patent law is built on the premise that inventors must be natural persons. This requirement has become problematic with the development of autonomous AI systems capable of generating patentable inventions without direct human input. The DABUS (Device for the Autonomous Bootstrapping of Unified Sentience) case epitomizes this issue. Stephen Thaler, DABUS produced two patentable inventions: a food container with fractal surfaces for improved grip and heat transfer, and a flashing light device for emergency signaling.

Dr. Thaler listed DABUS as the inventor in patent applications filed in multiple jurisdictions, including the European Patent Office (EPA), the United States Patent and Trademark Office (USPTO), and the UK Intellectual Property Office (UKIPO). Each of these offices rejected the applications, citing the requirement that only natural persons can be recognized as inventors. Legal authorities argued that patent systems were never intended to grant inventorship status to machines, regardless of the originality or utility of their outputs.

Recent clarification from the German Federal Court of Justice (BGH) in this case reaffirmed, however, that, under current German patent law, only natural persons can be designated as inventors.

The BGH’s ruling aligns with international patent law practices, emphasizing that the legal right of inventorship is reserved exclusively for natural persons. The court underscored, however, that while AI-assisted inventions can be patented, the inventive activity must be attributed to a human who has played a decisive role in the invention process, such as initiating or guiding the AI-driven innovation process.

While these decisions upheld existing legal norms, they sparked widespread debate among legal scholars, policymakers, and tech industry leaders. Legislative reform in this area remains a contentious topic, with proposals ranging from granting AI systems legal inventorship status to recognizing their developers as default inventors.

1.2 Disclosure and the ‘Black Box’ Problem

Another significant challenge in patenting AI innovations is the issue of technical disclosure, particularly when AI systems operate as ‘black boxes.’ Many AI models, especially those based on deep learning and neural networks, process data in ways that are difficult—even for their developers—to fully explain. This lack of transparency conflicts with the European Patent Convention’s (EPC) requirement that patent applications disclose an invention in sufficient detail to enable a person skilled in the field to reproduce it.

See German Federal Court of Justice decision, Case No. X ZB 5/22, from June 11, 2024.

The legal dilemma is that while AI-generated outputs may meet the criteria of novelty and inventive step, the underlying processes that created them may be too complex or opaque to satisfy disclosure requirements. For example, AI-driven drug discovery models often generate potential pharmaceutical compounds through unexplained pattern recognition and prediction, leaving gaps in the traditional disclosure narrative required by patent law.

1.3 Patent Eligibility of Algorithms and Mathematical Models

2 Copyright and AI-Generated Works: A Comparative Perspective

AI-driven creativity has also raised significant questions about copyright law. This legal vacuum contrasts sharply with emerging international legal developments, notably in China, EU and Germany in particular.

2.1 The Chinese Court Decision: Recognizing AI-Created Art

In a landmark ruling, the Internet Court of Beijing recognized copyright protection for an AI-generated image created using the generative AI model Stable Diffusion. In this case, the creator, Mr. Li, generated an artistic image by adjusting prompts, parameters, and aesthetic settings within the AI model. When a blogger named Liu used the image without permission, the court ruled in favor of Mr. Li, stating that his creative input met the legal standard for originality.

Beijing Internet Court decision, in Li v. Liu (2023) Jing 0491 Min Chu No. 11279, November 2023

The decision highlighted that although the AI system generated the image, Mr. Li’s creative input, including adjusting prompts, configuring parameters, and selecting aesthetic settings, constituted sufficient human involvement to meet the legal standard of originality under Chinese copyright law. The court emphasized that while the AI system executed the image creation, Mr. Li’s role in directing the creative process established him as the legitimate copyright holder. This ruling reflects an evolving legal view that recognizes human-guided AI creations while stopping short of extending copyright protection to fully autonomous AI works.

2.2 Comparative Legal Outlook

While the Chinese decision supports a more inclusive view of AI-generated content, the EU maintains a stricter interpretation centered on human authorship. Legal scholars have suggested creating new IP categories for AI-generated works or adapting existing copyright frameworks to reflect the collaborative nature of human-AI creative processes.

2.3 Legal Boundaries of AI Training: The Landmark Decision of the Hamburg Regional Court

The growing intersection of AI and copyright law has led to one of the most significant legal decisions in Europe concerning the use of copyrighted material for AI training.

In a landmark ruling, the Regional Court of Hamburg addressed the legality of using copyrighted photographs in AI training datasets under German and EU copyright law. In its judgment the Hamburg Regional Court ruled that using copyrighted photographs for AI model training is permissible under § 60d German Copyright Act (Urheberrechtsgesetz, UrhG) if used for scientific purposes. The court found that reproducing images for AI training aligned with copyright exceptions for Text and Data Mining (TDM), provided the use was non-commercial and research-focused.

Hamburg LG, case No. 310 O 227/23, of September 27, 2024.

The decision has far-reaching implications, especially for open-source initiatives that depend on publicly available datasets. By confirming that TDM exemptions apply to AI training datasets, the court reinforced the legal acceptance of using copyrighted materials in AI model development, provided the use is for scientific research and adheres to copyright law limitations. This development could encourage broader AI research collaborations and reduce the legal risks associated with training data collection.

The Hamburg court’s decision also referenced the EU AI Regulation of 2024, which explicitly allows the use of datasets for AI training purposes. This alignment between national and EU-level legal frameworks enhances legal certainty for AI developers operating across multiple jurisdictions. By acknowledging the EU AI Regulation, the court supported a harmonized approach to AI-related copyright challenges, making the ruling relevant beyond Germany’s legal boundaries.

Despite its groundbreaking nature, the ruling left one critical issue unresolved: the legal status of usage restrictions expressed in non-technical formats. The court considered whether copyright holders must declare usage reservations in machine-readable formats like robots.txt or if statements made in natural language are legally binding. While acknowledging the complexity of this issue, the court refrained from issuing a definitive ruling, leaving this question open for future legal interpretation.

[1] Beijing Internet Court decision, in Li v. Liu (2023) Jing 0491 Min Chu No. 11279, November 2023

3. The EU AI Act: Legal Framework and Implications

The EU AI Act, scheduled to take effect in 2026, aims to regulate AI systems based on their potential risks. Businesses developing such systems must comply with comprehensive technical and legal standards, including detailed disclosures of their models and decision-making processes.

The EU AI Act introduces a risk-based regulatory model, categorizing AI systems based on their potential societal impact. Businesses developing AI technologies must meet comprehensive legal and technical standards, including model transparency, explainability, and robust documentation of decision-making processes. These requirements have direct implications for IP law, as they involve disclosing proprietary algorithms, data sources, and model designs, which are often considered trade secrets.

3.1 Transparency Requirements and IP Protection:

One of the Act’s central demands is that businesses disclose key aspects of their AI models, including the technical processes underlying automated decisions. This creates tension with IP frameworks, particularly regarding trade secrets and patent law, where disclosure risks exposing valuable proprietary information. The challenge lies in balancing regulatory transparency with the protection of business-critical IP assets, a tension that underscores the need for harmonized legal definitions and procedures.

3.2 Accountability and IP Compliance

The EU AI Act holds businesses accountable for AI system outcomes, requiring clear attribution of responsibility across the AI supply chain. This principle overlaps with IP law, especially concerning software licenses, patents, and copyright ownership. Harmonizing these legal regimes across member states is essential to avoid conflicting interpretations of liability, authorship, and ownership in AI-driven innovation.

3.3 Standardized Legal Frameworks

By introducing uniform compliance obligations for AI systems, the EU AI Act helps reduce legal fragmentation among member states. This aligns with broader efforts to harmonize IP rights through institutions such as the EPA and the EUIPO. The Act’s provisions encourage consistent legal standards across Europe, reducing the regulatory burden for businesses operating across borders.

3.4 Innovation Incentives and IP Enforcement:

While the EU AI Act imposes regulatory obligations, it also supports innovation by clarifying legal uncertainties related to AI-generated works and patented technologies. A harmonized IP framework can foster a more predictable legal environment, enabling businesses to invest in cutting-edge technologies while securing enforceable IP protections throughout the EU.

II. Conclusion: Balancing Legal Tradition and Technological Innovation

Europe’s approach to intellectual property law reflects a delicate balance between legal tradition and the dynamic realities of technological innovation. While the EU has made significant strides in harmonizing IP law through institutions like the EPA, EUIPO, and the ECJ, the rise of AI poses unprecedented challenges that existing legal frameworks are ill-equipped to address.

As Europe continues to refine its legal approach, policymakers must weigh the potential for economic growth against the need for legal certainty and fairness. By doing so, the EU can maintain its competitive edge in the global innovation landscape while ensuring that its legal system remains robust, adaptive, and equitable.