There's a quote from Bernie Sanders that sounds more like philosophy than politics. "AI is being built on a public resource far more valuable than oil: humanity's accumulated knowledge, creativity, and labor."

It's technically precise. Most debates about artificial intelligence avoid following it to its logical conclusion because the implications are uncomfortable. Who is the legitimate owner of a system that learned to write by reading millions of texts that nobody sold to it? Who should receive dividends from a technology trained on the collective effort of generations?

Sanders has proposed legislation that attempts to give a concrete answer: a one-time tax collected in equity shares that would grant the U.S. government voting power and a seat on the boards of OpenAI, Anthropic, and xAI. It's not a bailout or content regulation. It's actual ownership participation.

What's interesting is that the argument doesn't come from Sanders alone. The labs themselves opened that door earlier. OpenAI has promoted forms of universal basic income. Anthropic talks about broadly distributing benefits. xAI has used the language of collective benefit. Each one, at different times, has suggested that AI's fruits should reach the general population. Sanders took those words and asked the obvious question: if that's what they believe, why not structure it with public participation that guarantees it?

The model he cites isn't invented. Norway manages a sovereign fund of more than two trillion dollars built on oil profits. It's not philanthropy. It's structural ownership. Every Norwegian owns, in an abstract but real way, a fraction of that fund. Alaska distributes annual cash dividends to its residents from the state's oil revenues. Both cases start from the same premise: when a natural resource belongs to everyone, the profits from its exploitation should also belong to everyone. Sanders argues that accumulated human knowledge is that resource and AI is exploiting it without paying rent.

I acknowledge the tension this type of proposal generates. It's not technically complicated. What makes it uncomfortable is that it touches something modern capitalism prefers to keep blurry: the line between what's built individually and what's inherited collectively. A language model didn't learn to write on its own. It absorbed Borges, García Márquez, millions of anonymous bloggers, entire forums, academic articles funded with public money. That training has authors. It's just that nobody asked them if they wanted to participate or if they expected some form of compensation.

The same question reappears every time a technology extracts value from a good that nobody built alone. The enclosure of common lands in eighteenth-century England displaced millions of people who had cultivated collectively for generations, all to grant legal titles to private owners. Radio and television used the electromagnetic spectrum—a public good—to build commercial empires. The internet was developed with government funding before becoming the foundation of some of the world's most valuable companies. In each case the central question was the same.

AI adds a deeper layer. It doesn't just use public infrastructure. Its raw material is humanity's complete cultural and intellectual output: every poem, every scientific article, every transcribed conversation, every labeled image, every shared line of code. All of that fed the models that are now worth hundreds of billions of dollars. It's not a metaphor. It's the concrete technical process by which these systems work.

The proposal faces real obstacles. OpenAI completed a transition to a for-profit structure that complicates any intervention in its governance. Anthropic maintains a form of public benefit without clear accountability mechanisms to society. xAI operates as a private company with the opacity that entails. Placing government on boards of directors isn't automatically positive. Governments can also be captured, prioritize surveillance over redistribution, or create bureaucracy without tangible benefit.

I'm still not clear if government equity participation represents the optimal path for benefits to reach those who unknowingly contributed. It could be. It could also end up legitimizing the current order by giving government a seat without altering who really sets the technological direction. From Mexico I see these dynamics differently, perhaps because here the gap between those who accumulate data and those who generate it feels more stark.

The question Sanders raises stands even if the solution he offers is imperfect. If humanity's collective knowledge is the resource on which history's most powerful AI is built, then the current model—where that repository is extracted without compensation to build private monopolies—carries a legitimacy problem that mere technological brilliance doesn't erase.

Norway's fund took decades to consolidate. Before that there were versions of the same debate about who had the right to profits from a resource that nobody planted. What's unclear is whether this time there will be enough time before the concentration of power makes any answer irrelevant.

Can government equity participation really guarantee that AI's benefits reach those who unwittingly built it, or does it simply transfer control from one type of elite to another?

Sources:

1. Sanders, Bernie. Proposed legislation on public participation in AI companies. U.S. Senate, 2025.

2. Government Pension Fund Global (Norges Bank Investment Management). Annual reports of the Norwegian sovereign fund.

3. Alaska Permanent Fund Corporation. Dividend history and fund overview.

4. Crawford, Kate. Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence. Yale University Press, 2021.

5. Doctorow, Cory. The Internet Con: How to Seize the Means of Computation. Verso Books, 2023.