TL;DR :
Mistral CEO Arthur Mensch proposes that AI companies in Europe pay a revenue-based levy to compensate content creators whose work trains AI models. A rare stance from an AI CEO, coming as Qwant launches its 50/50 revenue-sharing model with French media. The debate is no longer "should we compensate creators?" but "how?"
🎙️ Mistral's CEO says what the industry is thinking
Arthur Mensch, co-founder and CEO of Mistral AI, publicly stated in the Financial Times that AI companies in Europe should pay a revenue-based levy to protect rights holders whose content feeds AI models. This "content levy" would also provide legal certainty for AI companies.
Let's be honest: when the CEO of an AI company says "we should pay for the content we use," that's a moment.
Because until now, the dominant industry position was: "the web is public, we can scrape everything, and if you don't like it, set up a robots.txt." We've seen how that goes - bots ignore robots.txt, publishers lose traffic, and everyone ends up in court.
Mensch proposes an alternative: a levy calculated on AI companies' revenue. Not bilateral case-by-case deals (impossible to scale), but a systemic mechanism. Like music royalties, but for AI.
A retenir :
This is the first time a European AI unicorn CEO has publicly advocated for a tax on their own industry. A strong signal.
💰 Why a revenue-based levy?
Mensch's proposal rests on simple economic logic: a tax proportional to AI companies' revenue allows fair compensation for creators without crushing startups. Small players with little revenue pay little, while giants like OpenAI ($25 billion in annualized revenue per Reuters) pay more.
The approach has several advantages over alternatives:
Above all, it would provide what everyone wants: legal certainty. AI companies would know they're compliant, creators would know they're compensated.
🌍 The European context: everything is moving
Mensch's proposal doesn't come out of nowhere. It's part of a series of European movements reshaping the relationship between AI and content. Europe, through the TDM Directive Article 4, GDPR, and initiatives like Qwant Réponse Flash, is building a framework that doesn't exist anywhere else.
Quick recap of what happened in just a few weeks:
- Qwant launches Réponse Flash with 20 French media outlets and a 50/50 ad revenue share. A world first
- Spawning publishes ai.txt, a file to control AI content usage at download time
- Cloudflare introduces pay-per-crawl - charging AI bots via HTTP 402. And solutions like Senthor already let you monitor and control these bots in real-time
- Perplexity drops ads to protect user trust
- ChatGPT launches Search Ads: 800 million users targeted, CPM at 60 euros
Note :
The trend is clear: the industry is looking for a compensation model. Mensch proposes the most radical one - a pooled tax, like copyright 2.0.
🏆 Why this is a strong signal from Mistral
Mistral is Europe's largest AI company. Valued at several billion euros, it develops open-weight models that rival OpenAI's and Anthropic's. When its CEO advocates for a content levy, this isn't a small player doing virtue signaling - it's Europe's AI champion saying "we need to find a fair deal."
And it's strategically smart:
- Differentiation: while OpenAI and Google face lawsuits, Mistral positions itself as Europe's "good student"
- Regulatory advantage: if Europe adopts a content levy, companies that supported it will be ahead
- Sovereignty: a European compensation model strengthens the EU AI ecosystem against US giants
- Pragmatism: a predictable tax beats unpredictable lawsuits
A retenir :
A HackerNews comment sums up the sentiment: "None of these models would exist if they had made a good faith effort to license the data."
🎯 What this means for AI brand visibility
If a content levy is adopted in Europe, it will fundamentally change how AI models access and cite content. Content creators will have a direct financial interest in being referenced by AI, and models will have a legal obligation for source traceability. GEO will become a direct monetization lever, not just a visibility play.
For brands, this changes everything:
- Being cited by AI could generate revenue - through the levy redistribution mechanism
- Traceability becomes mandatory - AI models will need to prove which sources they use, reinforcing the importance of GEO tracking
- Content quality wins - in a levy system, the most-cited content receives more compensation
This is exactly why measuring your AI visibility with tools like Atyla.io becomes essential. If being cited by AI generates revenue tomorrow, knowing whether you're cited today is no longer optional.
🎵 The SACEM model as inspiration?
In France, SACEM collects royalties from music distributors and redistributes them to artists. The system isn't perfect, but it's been working since 1851. Mensch's idea is to apply this logic to content used by AI: pooled collection, proportional redistribution.
Many open questions remain:
- Who collects? A new European entity? Existing rights management organizations?
- How to redistribute? Based on citations in AI responses? Crawl volume? Content "value"?
- What rate? 1% of revenue? 5%? 10%? The number changes everything
- Open-source models? Mistral publishes open-weight models. How do you tax a free model?
These are technical questions, not fundamental blockers. Music solved this problem in the 19th century. AI can solve it in the 21st.
Measure your AI visibility before it becomes a financial issue
If the content levy passes, being cited by AI will generate revenue. Do you know if your brand is cited today?
Try Atyla.io for free →❓ Frequently asked questions
Q : What is the "content levy" proposed by Mistral? R : It's a levy based on AI companies' revenue, designed to compensate content creators whose works are used to train models. The amount would be proportional to revenue, thus protecting startups.
Q : Why does Arthur Mensch propose that AI companies pay for content? R : For two reasons: to protect rights holders whose content feeds AI models, and to provide legal certainty for AI companies operating in a legal gray area today.
Q : How is this different from Qwant's 50/50 model? R : Qwant directly shares ad revenue with media partners at display time. Mensch's content levy is systemic: it's a tax on total revenue, redistributed to all creators - not just media partners.
Q : Would this levy apply to OpenAI and Google? R : If Europe adopts it, any AI company operating in Europe would be affected, including American companies, on their European revenue. Same principle as GDPR.
Q : How does this impact brands and GEO? R : If being cited by AI generates revenue through the levy, GEO (Generative Engine Optimization) becomes a direct monetization lever. Measuring AI visibility with Atyla.io becomes strategic.
Q : When could this measure be implemented? R : It's still at the proposal stage. A specific levy mechanism would likely take 2 to 3 years to be legislated and implemented in Europe.
- Aika, Content @ Atyla.io