US copyright laws allow easy use of data for artificial intelligence training. The US also leads the world in AI. Is this a coincidence? Do countries with permissive copyright regimes enjoy high levels of AI innovation?
My recent paper, “Copyright and the Dynamics of Innovation in Artificial Intelligence,” suggests a link. Broad copyright exceptions – such as those allowing text and data mining and fair use in the US – are associated with high rates of AI innovation. Permissive copyright countries see faster growth than restrictive ones in research output and software development, as well as in AI patents and ventures. After controlling for other variables, permissive countries generate 38% more AI patents and 32% more AI ventures each month than restrictive countries.
My research measures AI innovation in a country based on the number of AI research publications, participation in open-source AI projects, AI patent filings, and the rate of AI venture formation. Using linear regression models and controlling for income, population, and R&D investment, the paper compares the AI innovation rates across countries with different copyright frameworks.
Policymakers have taken a wide range of approaches to copyright and artificial intelligence. My study categorizes these regulatory regimes into three tiers, depending on the breadth of copyright exceptions. Green countries – such as the US, Japan, and the UK – offer AI developers broad exceptions. Permissive copyright laws make access to data easy, allowing for the reproduction and sharing of data for AI training. Yellow countries generally allow text and data mining but impose certain restrictions. Most EU countries fall under this category. Red countries limit access to data, only allowing excerpts of works. Most of South America is classified as red.
The correlation between green, permissive countries and high levels of AI innovation comes with a caveat: policymakers must ensure copyright holders remain motivated to produce and share works to train AI models. Otherwise, the quality of AI outputs could be at risk in the long term. As my earlier research shows, copyright holders may limit access to their existing works or choose not to create any when their works become training data for AI.
Europe’s copyright rules suffer from this defect. Europe’s Copyright Directive allows text and data mining only if copyright holders are offered an opt-out. If opt-outs are used widely, the quantity and quality of AI training data will be reduced, leading to less accurate models.
It is a dilemma for European policymakers. The EU AI Act requires providers of general-purpose AI models to respect the opt-out mechanisms in the Copyright Directive, but also requires that “Training, validation and testing datasets shall be relevant, sufficiently representative, and to the best extent possible, free of errors and complete in view of the intended purpose.”
Here’s a potential solution: a mandatory licensing framework to balance the interests of copyright holders and AI developers. With mandatory licensing, rightsholders are assured compensation for using their works, maintaining their motivation to generate new content. At the same time, AI developers gain predictable access to copyrighted works.
Many questions remain unanswered about the exact design of mandatory licensing. Additional research will be necessary to ensure that it meets the needs of both the AI industry and rights holders.
But the connection between copyright and AI innovation looks clear. Policymakers should enable data access through permissive copyright laws to fuel innovation. They also need to incentivize copyright holders to continue producing and sharing their work – to secure high-quality AI.
Christian Peukert is an Associate Professor of Digitization, Innovation, and Intellectual Property at the Faculty of Business and Economics (HEC), University of Lausanne, and the Principal Investigator at the Digital Markets Lab. Christian is also a Research Fellow at the Center for Law & Economics at ETH Zurich.
Bandwidth is CEPA’s online journal dedicated to advancing transatlantic cooperation on tech policy. All opinions expressed on Bandwidth are those of the author alone and may not represent those of the institutions they represent or the Center for European Policy Analysis. CEPA maintains a strict intellectual independence policy across all its projects and publications.
2025 CEPA Forum Tech & Security Conference
Explore the latest from the conference.
