OpenAI Wants a Robot Tax
Arlo Gilbert ·
On Monday, OpenAI released a document called "Industrial Policy for the Intelligence Age." Thirteen pages. It proposes a robot tax on automated systems that displace workers. A nationally managed wealth fund, seeded by AI companies, that pays dividends to every American citizen. Government-incentivized 32-hour workweeks at full pay. Automatic economic tripwires that boost unemployment benefits when AI displacement crosses preset thresholds. Microgrants for displaced workers who want to start businesses.
Some of these are genuinely good ideas. The public wealth fund is modeled on Alaska's Permanent Fund, which has paid annual dividends from oil revenue since 1982. It's probably the most interesting structural proposal anyone in the industry has offered this year. The robot tax, first floated by Bill Gates in 2017, addresses a real fiscal problem: if AI replaces workers, payroll taxes that fund social programs erode. Taxing automated systems to replace that revenue isn't radical. It's math.
The 32-hour workweek proposal will get the most press and the least traction. It assumes companies will translate productivity gains into reduced hours rather than reduced headcount. Every prior wave of automation suggests the opposite. OpenAI seems to know this. The paper also proposes automatic stabilizers for when displacement happens anyway. It holds both ideas at the same time without acknowledging the tension between them.
The substance is worth reading. But the timing is the story.
OpenAI closed a $122 billion funding round on March 31 at a post-money valuation of $852 billion. Sam Altman is pushing for an IPO as early as Q4 2026. If it happens anywhere near current valuations, OpenAI will be one of the most valuable public companies on earth.
The paper landed six days after that fundraise closed. The IPO is months away. Thirteen pages that read like policy but function like positioning.
The biggest risk to OpenAI's IPO isn't technical. It isn't regulatory. It's narrative. If the public story about AI in late 2026 is "tech companies worth a trillion dollars are automating away middle-class jobs," going public is a hard sell. If the story is "the AI company that proposed taxing itself to share the gains," that's a different conversation entirely. OpenAI's paper doesn't commit the company to anything. It proposes that governments should consider things. But it puts OpenAI on the record as the company that thought about the workers.
In June 1889, Andrew Carnegie published "The Gospel of Wealth" in the North American Review. His argument: industrialists who had accumulated enormous fortunes had a moral obligation to redistribute them for the public good. Fund libraries. Build universities. Give back.
The essay was eloquent. It was widely praised. Carnegie also ran steel mills where workers pulled 12-hour shifts, six days a week, at wages that barely covered rent. Three years after publishing the essay, he authorized 300 Pinkerton agents to break the Homestead Strike by force. Seven workers died.
Carnegie meant what he wrote. He also benefited enormously from having written it. "The Gospel of Wealth" shifted the conversation about industrial barons from "these people are exploiting workers" to "these people are thinking deeply about their responsibility to society." The essay didn't change Carnegie's labor practices. It changed how people talked about them.
I'm not equating the two. OpenAI isn't running steel mills. But the structural pattern is old enough to recognize. When the entity proposing redistribution is also the entity being redistributed from, the proposal does two things at once. It's policy. It's also positioning.
Read the paper carefully and notice what's absent. It doesn't name any specific regulation OpenAI would support being applied to itself. It doesn't propose mandatory auditing standards for AI companies, enforceable transparency requirements on training data, or binding reporting mandates. It proposes taxes and wealth funds in the abstract. Governance frameworks and international coordination in the general. The paper is a blueprint for what governments should do around AI companies. It is not a blueprint for what AI companies should allow governments to do to them.
OpenAI frames the whole thing as exploratory. "A starting point for discussion," they write, "that we invite others to build on, refine, challenge, or choose among through the democratic process." That's fine language. It's the kind of language that creates the appearance of policy engagement without the weight of policy advocacy.
None of this makes the ideas bad. The wealth fund is smart. The automatic stabilizers are more honest than pretending displacement won't happen. Even the robot tax, clunky as the mechanics would be, grapples with a fiscal reality most AI companies won't touch.
But good ideas published at convenient moments, by companies with enormous financial stakes in how those ideas land, deserve to be read with both eyes open. OpenAI wrote thirteen pages about sharing the wealth. It filed zero pages of legislation.