By Douglas Noonan
I still remember a dinner years ago with Michael O’Hare, in the aftershock of Napster. The music business looked broken. Tower Records was going the way of the dodo. And O’Hare made an argument that sounded radical only because everyone else was still bargaining with reality: if the cost of making one more digital copy is basically zero, then the price of one more digital copy wants to be very low … like zero.
That idea has stayed with me ever since. It was stubbornly timeless.
Arthur Mensch, the CEO of Mistral, recently reopened that argument in public. His proposal is to levy commercial AI providers operating in Europe, including foreign firms, and send the proceeds into a central fund for new content creation and support for the cultural sector. The trade is straightforward: money for culture in exchange for legal certainty for AI developers, while direct licensing would still remain for uses beyond model training.
Let’s be clear: this is not philanthropy. Mistral wants a more workable route through Europe’s copyright thicket. Current EU rules already tie AI providers to copyright compliance and an opt-out text-and-data-mining regime, and policymakers are visibly unhappy with how that system is functioning. The European Parliament has recently called for fair remuneration, more transparency around training data, and sector-specific licensing markets, while explicitly resisting a flat-rate global license as the complete answer.
But Mensch’s self-interest does not make the diagnosis wrong. Sometimes it makes the diagnosis sharper.
The important point is not merely that AI companies should pay creators. Lots of people think that. The important point is that an AI CEO is effectively admitting that copy-by-copy control is becoming a clumsy foundation for financing cultural production in a digital world.
That is the part the arts world has been circling for twenty-five years. Timeless. And timely.
Music did not “solve” this problem so much as route around it. Instead of selling us one scarce copy at a time, the business moved toward subscriptions, catalog access, pooled revenue, and formula fights. Streaming is now the center of recorded music’s revenue structure, not the edge case. Globally, streaming now accounts for roughly 70% of recorded-music income.
That shift matters because it relocated the scarcity. The scarce thing is no longer the copy. It is attention.
And this is not just a music story. We’ve already been arguing that distribution costs are collapsing, that the arts economy is increasingly digital and subscription-like, and that time and attention are becoming the scarce resources that matter most. This has been happening across much of the arts-and-culture space for decades. Once you take that seriously, AI stops looking like a weird outside invasion. It looks more like an accelerant. It pushes more parts of culture into the same uncomfortable economic territory that music reached first.
Now, not all art suddenly has zero marginal cost. Live performance still has musicians, actors, stagehands, electricity bills, and seats that only one body can occupy at a time. The performing arts do not escape cost disease because Claude agents can sing, write, and draw all day long.
But more and more of the arts economy now runs through digitized artifacts, archives, clips, images, text, metadata, recommendation systems, and generative reuse. In that world, the old fantasy that copyright alone will neatly connect every use to every payment looks less like policy realism and more like nostalgia.
So I get why Mistral is pitching for a levy.
But a levy is only half a policy.
The real question is how the money gets allocated. That was the hard question in the Napster era, and it is the hard question now. On what basis do we distribute pooled compensation? Scrapes? Clicks? Listening time? Prompt influence? Output resemblance? Traffic diverted? Cultural impact? Need? Diversity? Local language protection? New entrants versus superstars?
This is where the argument gets interesting enough for arts policy.
Because once we stop pretending that scarcity at the copy level is doing all the work, we have to confront the fact that we are designing institutions. We are choosing formulas. We are deciding whether the future of culture doubles down on winner-take-all dynamics or deliberately tilts toward a broader ecology of creation.
And that is not a technical footnote. That is the whole game.
There is precedent for this kind of move. Private-copying compensation regimes emerged in many places because it was simply too impractical to license and monitor huge numbers of tiny acts of copying one by one. WIPO’s global survey makes that logic explicit. AI is obviously not the same thing as home taping. But the institutional rhyme is hard to miss. When individual permission becomes too administratively expensive relative to the behavior itself, levy logic starts to look less eccentric and more inevitable.
So here is the provocation.
Maybe the future of arts funding lies less in defending artificial scarcity and more in building robust ways to finance creation amid abundance.
That will sound like surrender to some people. It will sound like taxation to others. Many will resist change. Reasonable disagreements on optimal allocations will abound. Good. But let’s bravely and honestly tackle the challenge (and opportunity!) before us.
The price of access to culture has been drifting downward for decades. AI did not create that history. It’s accelerating it.
Napster was the warning shot. Streaming was the dress rehearsal. AI is the moment we decide whether we can let culture become more accessible without letting creators become more disposable.
If the Mensch proposal is useful for anything, it is this: it forces us to stop pretending that the old model is merely under pressure. For a growing share of culture, it may be the old model itself that no longer fits.
The price of culture wants to be zero.
The real question is whether we are finally ready to build a system that pays for creation anyway.