EU, UK, and France Chart Divergent Paths on Copyright Regulation for Generative AI

On 10 March 2026, the European Parliament adopted a Resolution titled "Copyright and Generative Artificial Intelligence – Opportunities and Challenges," signaling a strong political demand for urgent legislative action to address copyright issues arising from generative AI. Shortly thereafter, on 18 March 2026, the UK Government published its Report on Copyright and Artificial Intelligence, taking a more cautious stance by refraining from immediate legislative changes and emphasizing further evidence gathering. Meanwhile, on 8 April 2026, the French Senate advanced a bill introducing a rebuttable presumption concerning AI providers' use of copyrighted works, marking a notable national development in AI copyright enforcement.

These three texts confront the shared challenge of adapting copyright law to generative AI but diverge significantly in approach and tone. The European Parliament’s Resolution advocates for a prescriptive, rights-holder-centric framework with enforceable transparency and licensing obligations. In contrast, the UK Government favors a flexible, market-driven approach, avoiding new copyright legislation or regulatory oversight at this stage. The French Bill, narrower in scope, proposes a distinct evidentiary mechanism to protect rights holders by presuming AI use of copyrighted works when plausible indications exist.

The EU Resolution critiques existing copyright legislation, including the Directive (EU) 2019/790 on copyright in the Digital Single Market, as insufficient for licensing copyrighted material used in AI training. It calls on the European Commission to urgently propose supplementary legal frameworks to address these gaps. Conversely, the UK Report rejects earlier proposals for a broad copyright exception with opt-out provisions, citing strong opposition from creative industries. Instead, it advocates for continued monitoring of copyright impacts on AI development, exploration of targeted approaches, and observation of litigation and international developments before considering legislative reform.

Transparency emerges as a key theme with contrasting treatments. The EU Resolution mandates full transparency and source documentation from AI providers deploying general-purpose models in the EU market. This includes detailed disclosure of copyrighted works used in training, inference, retrieval-augmented generation, or fine-tuning. It further requires web crawlers to identify themselves and obliges AI providers to maintain comprehensive records of scraping activities. The Resolution clarifies that information about third-party content used by AI providers does not constitute a trade secret under EU law.

The UK Report acknowledges strong stakeholder support for transparency—over 90% of consultation respondents favored disclosure—but stops short of imposing mandatory transparency requirements. Instead, it proposes collaboration with industry and experts to develop best practices on input transparency, enabling rights holders to assert their rights effectively. The UK explicitly declines to establish new regulatory bodies or impose regulatory duties on existing agencies regarding transparency compliance.

On licensing and remuneration, the EU Resolution emphasizes that rights holders must retain full control over digital use of their content for AI training, supported by transparency and effective exclusion mechanisms. It proposes that the European Union Intellectual Property Office (EUIPO) act as a trusted intermediary to manage and catalogue opt-outs, and encourages voluntary sector-specific collective licensing agreements to foster a fair and operational licensing market. The Resolution opposes any blanket licensing schemes based on flat-rate payments.

In contrast, the UK Report adopts a less interventionist position, citing insufficient evidence to justify government interference in the licensing market. It relies on the existing copyright framework and an emerging voluntary licensing market, noting positive global trends in licensing deals. The UK also rejected the opt-out exception model similar to the EU’s approach, due to practical concerns expressed by both creative industries and AI developers.

The EU Resolution asserts extraterritorial application of EU copyright law to all generative AI models placed on the EU market, regardless of where training occurred. It proposes prohibiting non-compliant models from entering the EU market, representing a robust assertion of jurisdictional reach. The UK Report recognizes that most frontier AI models are trained outside the UK, primarily in the United States, but adopts a cautious approach. It declines to amend copyright law concerning systems developed abroad and relies on existing legal interpretations, such as the Getty Images v Stability AI judgment, which found insufficient evidence of UK-based infringement for AI training conducted overseas. The UK plans to monitor this issue, engage in bilateral discussions, and pursue international alignment through forums like the World Intellectual Property Organization (WIPO).

Regarding evidentiary burdens, the EU Resolution recommends a rebuttable presumption that copyrighted works have been used for training any generative AI model placed on the EU market if transparency obligations are unmet. When rights holders prevail on this basis, AI providers would bear all reasonable and proportionate costs, shifting the burden of proof in favor of rights holders.

The French Senate’s bill, adopted on 8 April 2026 but still subject to parliamentary approval, introduces a new Article L. 331-4-1 into the French Intellectual Property Code. This provision establishes a rebuttable presumption in civil proceedings that a copyrighted work has been used by an AI provider if an indication related to the AI system’s development, deployment, or output makes such use plausible. Unlike the EU’s presumption, which is triggered by transparency non-compliance, the French presumption applies more broadly and retroactively to pending cases.

The UK Report does not propose a similar presumption, relying instead on the existing civil enforcement framework, which it considers generally effective and adaptable to AI-related challenges. It plans further work to identify enforcement barriers, particularly concerning small and medium-sized enterprises and individual creators. The UK also highlights the potential clarifying role of the Getty Images v Stability AI appeal on secondary infringement for AI models trained abroad and explicitly rejects creating a new regulator for AI-copyright matters.

On the status of wholly AI-generated content, the EU Resolution confirms that such content not meeting established copyright criteria should be ineligible for protection and clearly designated as public domain. It also calls for mandatory labeling of exclusively AI-generated content and measures to combat deepfakes.

The UK Report reaches a similar substantive conclusion, proposing to remove the UK’s unique copyright protection for wholly computer-generated works, which has existed since 1988, due to minimal evidence of active use or economic impact. However, the UK diverges on approach by favoring voluntary industry collaboration to develop best practices on labeling AI-generated content rather than imposing mandatory labeling requirements, while continuing to monitor international developments.

These recent developments underscore a rapidly evolving regulatory landscape for generative AI and copyright. The EU is advancing a comprehensive, rights-holder-focused framework with enforceable transparency, licensing, and evidentiary provisions. The UK is adopting a measured, evidence-based approach, emphasizing existing legal frameworks and voluntary industry cooperation. France is pioneering a national evidentiary presumption to aid rights holders in civil enforcement. Companies operating in this space must navigate these divergent regimes carefully, anticipating forthcoming EU legislative proposals and ongoing UK and French policy evolutions amid increasing litigation and international dialogue.