TLDR: A new report commissioned by the European Parliament’s Committee on Legal Affairs (JURI) highlights a critical ‘legal mismatch’ between existing EU copyright law and the large-scale data training practices of generative AI models. Authored by Professor Nicola Lucchi, the July 2025 report calls for urgent reforms, including a shift to an ‘opt-in’ regime for using copyrighted works in AI training, fair remuneration for rightsholders, and clearer rules for AI-generated content to ensure a balance between innovation and creator rights.
Brussels, Belgium – A comprehensive report, ‘Generative AI and Copyright: Training, Creation, Regulation’ (PE 774.095), commissioned by the European Parliament’s Committee on Legal Affairs (JURI) and authored by Professor Nicola Lucchi, has delivered a stark assessment of the European Union’s current copyright framework, deeming it ill-equipped to handle the complexities introduced by large-scale generative Artificial Intelligence (AI) data training. Published in July 2025, the report emphasizes the urgent need for targeted reforms to protect copyright holders and ensure legal certainty in the evolving digital landscape.
The core challenge identified by the report lies in the massive ingestion and analysis of billions of copyrighted works required for training generative AI models. This process, involving acts of reproduction under Article 2 of Directive 2001/29/EC (the ‘InfoSoc Directive’), often occurs without explicit authorization from rightsholders. The report notes that even fragments of works can enjoy copyright protection if they reflect the author’s originality, as confirmed by the CJEU in Infopaq International A/S v. Danske Dagblades Forening (C‑5/08).
Existing Text and Data Mining (TDM) exceptions under the Directive (EU) 2019/790 (the ‘DSM Directive’) – Article 3 for scientific research and Article 4 for other purposes (with opt-out) – are deemed ‘ill-suited to the realities of GenAI.’ The report highlights that current opt-out mechanisms, such as metadata or robots.txt, are technically insufficient, and fragmented national transpositions of Article 4 create significant legal uncertainty for stakeholders.
Furthermore, the report scrutinizes the contribution of Regulation (EU) 2024/1689 (the ‘AI Act’). While the AI Act imposes transparency obligations on general-purpose AI model providers (Articles 53 and 55), it fails to resolve the fundamental question of whether mass reproductions of copyrighted works for AI training require prior authorization. The report strongly advocates for a shift to an ‘opt-in’ regime, mandating developers to obtain explicit permission before incorporating protected works into their training datasets.
Addressing the legal status of AI-generated content, the report distinguishes between autonomously generated AI works and AI-assisted works. Drawing on CJEU jurisprudence (e.g., Painer (C‑145/10) and Levola Hengelo (C‑310/17)), it asserts that copyright protection requires a ‘work’ to result from ‘the author’s own intellectual creation’ reflecting ‘free and creative choices.’ Consequently, works autonomously generated by AI, lacking meaningful human intervention, should be excluded from copyright protection. For AI-assisted works, the report calls for a nuanced assessment of human input, clarifying that merely entering prompts is insufficient for authorship, but actions like selecting outputs, iterating prompts, editing, or creatively combining AI-generated content may qualify. To prevent divergent national approaches, the report recommends adopting a directive specifying originality criteria for AI-assisted works, issuing EUIPO guidelines, and explicitly rejecting a sui generis right for AI-generated content to maintain the coherence of the EU copyright framework.
The report also raises concerns about the potential ‘dilution of human creativity’ if hybrid works receive broad protection and highlights risks to moral rights (Article 6bis of the Berne Convention) when AI manipulates or derives from human-authored works without consent.
To pave the way for a robust EU copyright reform, the report proposes a clear roadmap:
Opt-In Regime: Establish a European register of permissions and require traceability technologies for AI training data.
Fair Remuneration: Introduce extended collective licensing systems or statutory remuneration schemes, potentially inspired by private copying levies, to bridge the value gap between AI developers and creators.
Strengthened Governance: Create a dedicated ‘AI and Copyright’ unit within the European AI Office to oversee compliance audits and coordinate regulatory efforts.
Also Read:
- European Union Advances Landmark AI Regulation with New Guidelines and Imminent Enforcement
- Entertainment Giants File Landmark Copyright Lawsuit Against AI Image Generator Midjourney
In its conclusion, the report frames its recommendations as a ‘call to action,’ emphasizing the need to balance technological innovation with the preservation of cultural diversity. As stated in the report, ‘The European Union has an opportunity to become a model for balanced governance, reconciling technological innovation with the preservation of cultural diversity.’ The findings underscore the critical juncture at which EU policymakers find themselves, tasked with shaping a legal framework that supports both creative industries and the burgeoning AI sector.
