Bill Gates Wants to See a Robot Tax and 'Human Reserved' Jobs

Bill Gates published a 6,000-word essay on August 26, 2026 calling for two structural interventions in how societies govern AI: a tax on robots and automated AI token usage to slow the economic pressure to replace human workers, and a formal category of jobs — "human reserved" — that AI would be legally barred from performing. The essay, titled The turbulent AI era is here. The choices we make are critical, is Gates' first extended writing on AI in three years, and it lands at a moment when the policy vacuum around labor displacement has become impossible to ignore.

What's Converging

Public anxiety about AI and jobs has been building for years, but it has recently crossed from sentiment into measurable behavioral data. A Pew Research study released the week before Gates' essay found that 71% of American adults now believe AI will lead to fewer jobs over the next two decades — up from 64% in 2024. Only 5% expect AI to create more jobs. What the Fortune coverage adds that other outlets underplayed: younger workers are just as worried as older ones, with 73% of young people believing they will have fewer career opportunities because of AI over the next 20 years. This cuts against the familiar reassurance that younger workers will adapt and retrain; at scale, they are not buying it.

Meanwhile, the policy apparatus has lagged badly. The Irish Times frames Gates' concern sharply: he argues that what AI demands from governments is a response larger in scope than the institutional reorganization the United States undertook after September 11, 2001. That's a concrete historical benchmark, not vague alarm. And yet the legislative response in most jurisdictions remains fragmented — a regulation here, a voluntary commitment from labs there. The governance architecture Gates is describing simply does not exist yet.

Into that gap, Gates is proposing not a wait-and-see posture but two specific mechanisms. That specificity is worth noting. Plenty of prominent technologists have warned about labor displacement without putting anything concrete on the table. Gates is doing something different.

The Specific Development

The first proposal targets a real asymmetry in the tax code that most coverage has mentioned but few have explained clearly. Under current rules, when a business hires a human worker, payroll taxes apply — a direct cost on top of wages. When that same business replaces the worker with a robot or automated system, the capital expenditure can often be written off immediately. Gates argues this creates a structural incentive for labor-shedding that has nothing to do with whether automation is the right choice for a given role; it's just cheaper on paper. His proposed robot tax would narrow that gap, with the revenue directed toward retraining programs and economic transition support. Fortune's reporting adds a detail the other outlets omit: Gates also raised taxing AI tokens specifically, meaning the per-unit cost of AI inference itself — a mechanism that would create a direct levy on the throughput of automation rather than just the hardware.

The second proposal is conceptually more ambitious. Gates wants to formalize a category of work called "human reserved" — jobs that AI would be prohibited from performing, either because the workers in those roles cannot realistically retrain, or because the human presence is itself the point of the job. The Guardian captures the distinction Gates draws most clearly through the nature reserve analogy: just as society designates land that could be developed but chooses to protect, "human reserved" work would be an active choice to preserve human roles even when AI substitution is technically feasible. He is explicit that this is not a claim about capability. The Irish Times quotes him directly on the health-care case: "Imagine a robot giving you the awful news that you have an incurable disease. There's no technical reason why it couldn't. Yet it shouldn't." He also cited, across multiple outlets, the medical professionals who cared for his father, who died of Alzheimer's disease six years ago, as an example of work that was "irreplaceably human."

Our read is that the "human reserved" framing is the more consequential of the two proposals — not because the robot tax is less important, but because it introduces a named, citable policy concept that regulators and litigants can actually work with. Once a category has a name and a rationale, it becomes possible to argue about its boundaries in court, in legislation, and in procurement contracts. That is a very different kind of policy lever than a tax rate adjustment.

Gates also raises what is arguably the hardest question: who decides what gets designated human reserved, and by what process? He explicitly acknowledges in the essay that this is unresolved, writing — as the Guardian notes — that it "is fair to wonder whether the world's institutions are up to the task of designing and implementing this new architecture." He is not pretending the governance problem is solved. Multiple reports indicate he frames both proposals as directional rather than prescriptive, expecting the specific mechanisms to be contested. He also notes pointedly that both ideas would directly reduce revenue for the major AI labs, which he suggests may explain why neither has received serious public discussion until now. That is a political claim, not just a technical one, and it will not land without controversy.

What's Likely Next

The immediate question is whether either proposal finds a legislative sponsor willing to take the political heat. The robot tax concept is not new — economists have circulated versions of it for years — but it has never gotten traction in the United States largely because business lobbying against any capital-expenditure restriction is well-organized and well-funded. Gates' essay gives the idea more mainstream cover than it has had, but cover alone does not move votes. Watch for whether any sitting legislator cites this essay in the coming weeks; that would be a signal the conversation is moving from think-tank discourse into actual committee consideration.

The "human reserved" framing faces a different set of problems. Drawing the line between a job that is prohibited from AI substitution and one that is merely regulated is enormously difficult, and the line will shift as AI capability improves. What's technically prohibitive today is trivial in three years. Gates acknowledges a phased approach — slowly bringing AI into some sectors over years or decades — but the governance structure to manage that phase-in does not exist anywhere. The next 30 to 90 days will tell us whether this essay is a one-news-cycle moment or the start of an organized effort to put these frameworks into specific bills or international agreements. The scale of the Pew numbers suggests the public appetite for intervention is there. Whether institutions can move fast enough to meet it is another question entirely.

Sources