The four-day workweek is a bribe

On April 6, OpenAI released a 13-page policy paper called Industrial Policy for the Intelligence Age. Among its proposals: a four-day workweek with no loss in pay, taxes on automated labor, portable benefit accounts, and a public wealth fund seeded by AI-driven growth. The pitch is simple. AI will make us so productive that companies can give workers a day off and still come out ahead. Everyone wins. It sounds generous. It sounds progressive. It sounds like the kind of thing a company worth hundreds of billions of dollars might say right before it displaces millions of jobs. Because that is exactly what it is.

The offer on the table

OpenAI's paper frames these proposals as "initial ideas" for navigating what it calls the transition to superintelligence. The language is careful, even deferential. "We invite others to build on, refine, challenge, or choose among these through the democratic process," the paper reads. The headline proposals include encouraging companies to pilot four-day workweeks, boosting retirement contributions, covering more healthcare costs, subsidizing childcare, and creating portable benefit accounts that follow workers across jobs. On the fiscal side, OpenAI suggests shifting taxation away from earned income toward corporate profits and capital gains, with special consideration for taxes on automated labor, sometimes called a "robot tax." Taken at face value, these are reasonable ideas. Some are genuinely interesting. But the framing reveals more than the proposals themselves.

A concession, not a solution

There is a long history of powerful institutions offering small comforts to forestall larger reckonings. During the Industrial Revolution, factory owners in Britain worked laborers for 14 to 16 hours a day, six days a week. When public pressure mounted, some offered marginally shorter hours or modest improvements to working conditions. These concessions were not born from generosity. They were strategic. Give just enough to defuse the anger, not enough to change the power structure. OpenAI's four-day workweek proposal follows the same playbook. It assumes a world where people still have jobs to work four days at. It assumes employers will voluntarily pass productivity gains to workers rather than pocket them. And it frames these assumptions as policy, not aspiration. The four-day workweek is an answer to a question nobody is really asking. The harder question, the one OpenAI's paper dances around, is this: when AI can do the work of ten people, who benefits from the output of those ten?

The Jevons paradox problem

In 1865, the economist William Stanley Jevons observed something counterintuitive about coal. As steam engines became more efficient, you would expect coal consumption to drop. Instead, it surged. Greater efficiency made coal cheaper per unit of useful work, which unlocked entirely new uses and drove total demand through the roof. This is Jevons paradox, and it applies directly to AI and labor. If AI makes a knowledge worker five times more productive, the naive assumption is that the worker can now do five days of work in four. But that is not what happens in practice. What happens is that the company realizes it can extract five times the output from one worker, or get the same output from a fraction of the headcount. The "efficiency dividend" does not flow to the worker as a day off. It flows to the company as margin. OpenAI's proposal quietly assumes companies will behave against their own financial incentives. It asks employers to convert productivity gains into leisure rather than profit. History suggests they will do the opposite. Satya Nadella invoked Jevons paradox himself when DeepSeek threatened to undercut OpenAI's pricing. "As AI gets more efficient and accessible, we will see its use skyrocket," he wrote, sharing the concept's Wikipedia page. He was right, but the implication cuts both ways. If demand for AI is insatiable, why would any company settle for less output just because the workweek got shorter?

The public wealth fund question

Of all OpenAI's proposals, the public wealth fund is the most structurally interesting, and also the most conveniently vague. The idea is straightforward: create a fund that gives every citizen a stake in AI-driven economic growth. The fund would invest in diversified, long-term assets tied to AI companies and the broader economy. Returns would be distributed directly to citizens. This is not a new idea. Sovereign wealth funds exist in Norway, Singapore, and Alaska. The concept of distributing AI-generated wealth more broadly has been floating around policy circles for years. Sam Altman himself has talked about variants of this since at least 2021. But the paper leaves the critical questions unanswered. Who seeds the fund? How much? Who manages it? What governance structure prevents it from becoming another vehicle for the same companies that created the displacement in the first place? As TechCrunch noted, OpenAI frames employer responsibilities like healthcare subsidies and retirement contributions as corporate obligations rather than government ones. This is a revealing choice. If automation eliminates your job, your employer-subsidized healthcare and retirement match go with it. Portable benefit accounts help, but they still depend on employer or platform contributions. They stop short of the government-backed universal coverage that would actually protect people displaced entirely. The fund sounds like shared prosperity. But without clear answers about control and governance, it risks becoming a PR mechanism, a way to say "we're thinking about this" without actually redistributing power.

The Singapore angle

Singapore is an interesting lens for this conversation. The country was built on a particular bargain: work hard, stay productive, and the system will take care of you. The government has invested heavily in workforce development, from SkillsFuture credits to national AI literacy programs. Singapore consistently ranks among the highest adopters of AI at work, with 68% of workers already using AI tools in their jobs. But Singapore's social contract has always been predicated on labor as the primary mechanism for dignity and security. CPF contributions, housing eligibility, healthcare subsidies, nearly everything flows through employment. When OpenAI talks about a four-day workweek, the implicit assumption is that work remains the organizing principle of life. For Singapore, where that assumption runs even deeper than in the West, the question is not just about how many days you work. It is about what happens to the entire social infrastructure when the link between labor and livelihood starts to fray. The IMF has flagged that Singapore is highly exposed to AI disruption precisely because of its large share of skilled knowledge workers. Women and younger workers face particular vulnerability. A four-day workweek does not address any of this. It is a Band-Aid on a structural wound.

What is actually being avoided

The real conversation OpenAI does not want to have is about ownership. When a factory automated its assembly line in the 20th century, the gains went to the factory owner. When AI automates knowledge work in the 21st century, the gains go to the companies that own the models, the compute, and the data. The worker who used to write the report, analyze the data, or draft the contract does not get a share of the surplus their replacement generated. They get a severance package if they are lucky, and a suggestion to "reskill." OpenAI's proposals tinker with redistribution at the margins. A shorter workweek here, a tax adjustment there, a wealth fund somewhere in the future. But none of them challenge the fundamental question: should the companies building AI that replaces human labor be required to share ownership of that capability with the public? This is not a radical question. It is the same question societies asked during every previous technological revolution. And every time, the answer was deferred until the concentration of power became too large to reverse without enormous disruption.

The proposals are not bad, they are insufficient

To be clear, nothing in OpenAI's paper is wrong in isolation. Four-day workweeks are good. Portable benefits are good. Taxing capital gains more heavily is good. A public wealth fund could be transformative if done right. But packaging these ideas as a policy framework for the "Intelligence Age" while simultaneously building the very technology that will make many of them irrelevant is, at best, premature optimism. At worst, it is a calculated move to shape the policy conversation on terms favorable to the companies doing the disrupting. OpenAI is not offering a new social contract. It is offering a severance package dressed up as vision. The four-day workweek is not a gift. It is a negotiating tactic, just enough comfort to keep the public from asking who really owns the future. The question is not how many days we work. The question is who owns the output when the work no longer needs us at all.

References