Mistral Proposes Revenue Levy on AI Companies to Compensate Creators
French AI developer Mistral called Friday for European AI companies to pay a levy on revenues to fund content creators and cultural industries, addressing a growing dispute over how generative AI models are trained on human-generated material.
Mistral CEO Arthur Mensch said operators should contribute "a revenue-based levy... reflecting their use of content publicly available online," according to an op-ed in the Financial Times. The company suggested a rate between 1.0 and 1.5 percent of revenues.
Proceeds would flow into a central European fund for new content creation and cultural sector support, Mensch wrote.
How It Would Work
The levy would apply equally to foreign AI companies operating in Europe, creating what Mensch called "a level playing field within the European market." Most major AI developers are based in the United States.
In return for paying, AI developers would gain legal certainty and protection from liability for training on publicly available materials, according to Mensch. The mechanism would not replace direct licensing agreements between data owners and AI firms.
The Copyright Problem
AI models train on vast amounts of text, audio, and video created by humans. This has prompted lawsuits and complaints from creators and copyright holders on both sides of the Atlantic.
The EU adopted AI regulation in 2024 requiring systems to respect copyright rules, but applying these rules to generative AI remains unsettled.
Mistral itself faced accusations last month of using copyrighted works including "Harry Potter" and "The Little Prince" to train its model, according to an investigation by French media outlet Mediapart. The company said it respects opt-out mechanisms and deploys safeguards, though some widely duplicated works are difficult to exclude entirely from training data.
Comparison to U.S. Approach
American AI companies operate under "extremely permissive regulatory contexts on copyright," Mistral's external affairs chief Audrey Herblin-Stoop said. Anthropic, a major U.S. AI developer, agreed in September to pay $1.5 billion to settle a class-action lawsuit by authors.
Mistral, valued at 11.7 billion euros ($13.5 billion), positions itself as Europe's challenger to U.S.-based AI giants with valuations in the hundreds of billions.
For creatives and content creators, the proposal represents one approach to ensuring compensation when their work trains commercial AI systems.
Your membership also unlocks: