Arthur Mensch, CEO of Mistral, proposed on March 20 that companies developing artificial intelligence (AI) models in Europe should contribute a revenue-based “levy” to support cultural industries. Mensch made the suggestion in an op-ed for the Financial Times during the 56th annual World Economic Forum meeting in Davos, Switzerland.
Mistral’s call for the levy stems from growing concerns surrounding the use of human-generated content in training AI models, which has sparked legal challenges from creators and copyright owners across Europe and the United States. The proposed levy would be between 1.0 and 1.5% of revenue, according to Mistral’s external affairs chief, Audrey Herblin-Stoop.
“Proceeds would flow into a central European fund dedicated to investing in new content creation and supporting Europe’s cultural sectors,” Mensch stated. He emphasized that the levy should apply equally to both domestic and foreign AI providers, ensuring a level playing field within the European market. This would mean that US-based companies operating in Europe would also be subject to the same regulations.
The European Union’s AI regulation, adopted in 2024, mandates compliance with copyright laws, but the specifics of how these rules apply to generative AI systems remain unresolved. Mensch argued that the levy would provide AI developers with “legal certainty,” protecting them from liability when training on publicly available materials online. However, he clarified that this measure would not eliminate the need for direct agreements between data owners and AI firms.
The suggestion comes against the backdrop of significant financial stakes in the AI sector. Mistral, valued at €11.7 billion, seeks to position itself as a formidable competitor to established AI giants in the United States, which have valuations in the hundreds of billions. According to Herblin-Stoop, these dominant players benefit from “extremely permissive regulatory contexts on copyright.”
In a notable case, the American AI company Anthropic agreed to pay $1.5 billion to resolve a class-action lawsuit involving authors, highlighting the contentious landscape of AI development and intellectual property rights.
Adding to the complexity, Mistral recently faced allegations of utilizing copyrighted works, including Harry Potter and The Little Prince, to train its AI model. The company responded to these claims by stating that it “respects the opt-out mechanisms and deploys safeguards” to prevent the inclusion of copyrighted material. Nonetheless, due to the popularity and extensive online presence of certain works, completely excluding them from training data poses a significant challenge.
Mensch’s advocacy for a revenue levy reflects an ongoing evolution in the dialogue surrounding AI regulation in Europe. As the industry continues to expand, the balance between fostering innovation and protecting intellectual property rights remains a contentious issue. The proposed levy could not only provide essential funding for cultural sectors but also establish clearer guidelines for AI developers navigating the complex landscape of copyright law.
See also
Perplexity AI CEO Aravind Srinivas: Google Search Outperforms Competitors, Including Comet
Jury Finds Elon Musk Liable in Twitter Fraud Case; Damages Could Reach $2.6 Billion
Mistral Proposes Revenue-Based Levy for AI Training Copyright in Europe
Google DeepMind Appoints Jasjeet Sekhon as Chief Strategy Officer Amid AI Competition
Microsoft Acknowledges Overreach, Reduces Copilot Integration in Notepad and Other Apps
