The Numbers Come First
Wikipedia turns 25 with no ads, no paywalls, and an average donation of roughly $10. Millions of donors give year after year to keep the world's largest digital encyclopedia running. That is the financial baseline Jimmy Wales described to Fortune — and it is the same baseline, he argues, that major artificial-intelligence companies are quietly exploiting at scale.
'It's not reasonable for our millions of donors, who contribute an average of around $10, to subsidize large technology companies,' Wales told Fortune. 'They are supporting our mission.'
The statement is blunt, and the math behind it is straightforward. Wikipedia's content is freely licensed — anyone can use, modify, and redistribute it, commercially or otherwise, at no cost. Wales confirmed that under the current licensing model, Wikipedia cannot charge AI firms for training data. What it can demand, he said, is that high-volume use be handled 'in a more structured and fair way, through systems we can manage,' rather than placing an unmanaged burden on infrastructure funded by small donors.
AI Gets the Data; Wikipedia Gets the Bill
The arrangement raises a clean free-market question: when a private company extracts value from a public-good resource, who bears the cost? In this case, the answer is the individual donor who clicked 'give $10' to keep knowledge free. Wales stopped short of calling for regulation, noting that 'most major AI companies are beginning to recognize that they need to be fair to Wikipedia' and that 'things are moving in the right direction.' But the direction of travel matters less than the speed.
On the broader question of AI and knowledge quality, Wales was measured rather than alarmist. He said he uses AI 'extensively' and finds it makes users 'more active, not more passive' — citing his own hobby programming as an example where AI helps him understand concepts when he asks for explanations and follows up with complex questions. The caveat: 'not everyone will use it that way.'
He also pointed to what he called a 'core problem' with large language models — hallucinations, the tendency of systems that predict the next most-likely word to produce confident-sounding errors. That structural limitation, he suggested, is precisely where Wikipedia's model of transparency, verification, and group responsibility retains an edge that AI has not yet earned.
'AI can generate answers,' Wales said. 'It still can't earn trust.'
What the Market Has Already Voted
Wikipedia's durability is itself a market signal. Twenty-five years without advertising revenue, without subscription pressure, and without the incentive to chase sensational headlines — and the platform remains the default starting point for human curiosity worldwide. That is not an accident of charity. It is the result of a governance model built on verifiable sourcing and distributed accountability, two things that scale with human effort rather than compute power.
The editorial read here is simple: free enterprise works best when the rules are clear and the costs are honest. Right now, the cost of training the world's most powerful AI systems is being partially offloaded onto a volunteer-and-donor network that was never designed to carry it. Wales is not asking for a government bailout or a Brussels directive. He is asking for fair dealing — the oldest principle in commerce. Capital rewards clear rules, and the AI industry's long-term credibility depends on whether it can honor them without being compelled to.
