One of the rewarding aspects of academic work is the occasional opportunity to contribute to a policy debate while it is still taking shape.
Writing my report for the European Parliament, The Economics of Copyright and AI: Empirical Evidence and Optimal Policy, was such an opportunity.
The debate about copyright and generative AI is moving exceptionally quickly. Policymakers must make decisions while the technology, markets and available evidence are still evolving. The debate is also often presented as a conflict between creators and technology companies: rights holders want to protect their work, while AI companies want access to training data.
The report gave me an opportunity to step back from this framing. From an economic perspective, the question is not only how revenues should be divided between creators and AI firms. It is also how copyright policy can preserve the benefits of AI while ensuring that people continue to produce the fresh, high-quality works on which future AI systems may depend.
I found the assignment particularly satisfying because the connection to policy was unusually direct. The report had been commissioned by the European Parliament’s Committee on Legal Affairs as legislators were considering how European copyright law should respond to generative AI.
The analysis brought together historical evidence from earlier waves of digitization, recent empirical research on creators’ responses to AI training, estimates of the value generated by generative AI, and a formal economic model. I used this framework to compare four policy approaches: a broad copyright exception, an exception with an opt-out, voluntary licensing, and statutory licensing.
My research suggested that a carefully designed statutory licence, combined with a modest royalty, could perform relatively well across a range of scenarios. It could preserve broad access to training data, reduce the transaction costs of negotiating with large numbers of rights holders, and provide compensation that supports incentives for future creation.
This is not an easy policy to implement. AI companies may prefer free access, while many rights holders understandably want to retain control over the use of their work. A statutory licence also raises difficult questions about royalty rates, administration and the distribution of payments. The report did not resolve all these questions. Its more limited purpose was to examine the economic trade-offs and identify an option that deserved closer consideration.
What happened next
The European Parliament did not follow the report’s central recommendation. Its March 2026 resolution (nicknamed the "Axel Voss report") favored voluntary licensing, collective agreements and mechanisms through which rights holders can reserve their rights. An article in the Journal of Intellectual Property Law & Practice by Bernd Justin Jütte and Paulina Ryszkowska later noted that Parliament’s position differed from the conclusions of the economic study.
It would therefore be difficult to claim that the report directly influenced EU policy. But there have been some signs that it contributed to the wider discussion.
When MEP Reinier van Lanschot explained why Volt had abstained from the final vote, he referred to the expert studies commissioned by Parliament. He argued that statutory remuneration offered a possible compromise and that the adopted approach risked strengthening the position of the largest technology companies and rights holders.
The report has also appeared in discussions outside the immediate EU legislative process. India’s Department for Promotion of Industry and Internal Trade cited it in its working paper, One Nation, One License, One Payment: Balancing AI Innovation and Copyright, when considering different licensing frameworks. The Indian paper developed its own hybrid proposal, and its timing means that one should be cautious about drawing conclusions concerning influence. Still, it was interesting to see the analysis being used in another jurisdiction.
Policy researchers and legal scholars have engaged with the report as well. A Brookings analysis of the emerging AI content-licensing market by Courtney Radsch, PhD drew on its discussion of fragmented licensing markets and the future supply of creative content. It supported statutory licensing as a possible starting point, while proposing a different version that would allow publishers to opt out and negotiate private agreements.
The Kluwer Copyright Blog article by Marianna Foerg presented the economic reasoning in some detail, while also pointing to an important limitation: making an economic case for statutory licensing does not solve the legal and institutional problems involved in putting such a system into practice.
The report has also attracted criticism. A report from the French Ministry of Culture’s copyright council by Alexandra Bensamoun took issue with what it saw as an overly consumer-oriented view of copyright. It argued that human creation should not be understood primarily in terms of satisfying consumer demand or supplying content to future AI systems. I do not fully agree with this reading of the report, but the criticism identifies a genuine difference in perspective. Economic welfare analysis cannot capture every cultural, moral or legal value associated with creative work. At the same time, I believe it can help make some of the costs and trade-offs of different policy choices more explicit.
A broader view of policy relevance
This experience has been a useful reminder that research rarely enters policy in a straightforward way. A study may not be translated into legislation. Its arguments may instead be adopted in part, modified, contested or used to clarify disagreements.
The discussion surrounding the report reflects a broader uncertainty about what copyright policy should seek to achieve in the age of AI. Relevant objectives include protecting creative autonomy, compensating rights holders, sustaining cultural production, enabling the development of useful AI systems, and preserving the benefits these systems provide to users. These objectives overlap, but they do not always point toward the same policy.
It is gratifying to see that some policymakers, researchers and practitioners have read it, discussed its conclusions and, in several cases, disagreed with them. For an academic, that feels worthwhile. The experience reinforced my belief that engaging with policy debates matters. Research can contribute modestly by organizing the available evidence, clarifying the choices and helping to make the underlying trade-offs more visible.