Bill Gates spent five decades arguing that technology, on balance, makes the world better. On Wednesday morning he published a 5,784-word essay arguing that governments may need to stop it — deliberately, by law, in specific occupations — to keep people employed.
The essay, titled "A turbulent AI era and critical choices to make" and posted to Gates Notes on August 26, lays out three proposals. Two are conventional: new national coordinating bodies for AI, and an international institution modeled on nuclear weapons inspections, aviation rules and the ozone treaties. The third is not. Gates proposes a category of work he calls "Human Reserved" — jobs machines will be fully capable of doing, that society decides to keep for people anyway.
The analogy he reaches for is a nature reserve: land that could be developed, roads and buildings and profit attached, but left alone because the loss would be too great. Applied to labor, the logic produces uncomfortable examples. A machine delivering the news that you have an incurable disease is one Gates raises directly. "There's no technical reason why it couldn't," he writes. "Yet it shouldn't." The idea came partly from watching the caregivers who looked after his father through Alzheimer's — work he calls "irreplaceably human."
The 40 Percent Question
In an interview with Axios published alongside the essay, Gates put a number on it. Asked how much of the labor market could plausibly be walled off this way, he said: "In a very extreme form of it," roughly 40 percent of jobs could initially be reserved for humans. "But that's as high as I can get." He told GeekWire he worked the arithmetic through with Claude, testing whether shorter workdays and earlier retirement could stretch the remaining work far enough. Even 40 percent proved harder to construct than he expected.
Gates named child care and jury service as clear cases. Education and health care would be mixed — "you wouldn't want to completely ban" AI in either, he told Axios; instead a profession stays protected while its workers use AI to do more. He described two flavors of reservation: permanent, and temporary. The second targets people who realistically cannot retrain — manufacturing workers in the last stretch of their careers, for instance — with AI phased into an occupation over years rather than dropped in at once.
The essay pairs this with a revived version of a proposal Gates first floated in February 2017: taxing robots, updated here to include AI tokens. His argument is structural rather than punitive. Hire a person and you pay payroll taxes; buy a robot and you deduct the cost. "The tax system nudges you toward replacing people with machines," he writes. He concedes the tax is not economically efficient, and told Axios the ask is enormous: "I'm talking about a change to the tax system that's greater than any in my lifetime. At a time when politics is more polarized than any time in my lifetime."
Gates also said the private mood in AI labs diverges sharply from the public one. Executives who voice concerns, he told GeekWire, get told: "Hey, you're hurting our PR while we're trying to raise trillions of dollars." He would stake his reputation on the forecast: "I'd love to be convinced that I'm wrong about the job market, but I'm not."
Why It Matters
The case for reserved work is that retraining, the default policy answer since the 1980s, assumes displaced workers can move somewhere the machines are not. Gates's premise is that AI advances broadly enough that the destination disappears. If that holds, a community college voucher is not a bridge to anywhere. Reserving occupations also encodes a judgment markets do not price: that being told you are dying, or having your child cared for, carries value from the humanness of the actor, not just the output. Several countries already do versions of this quietly — licensure rules, minimum staffing ratios in nursing homes, requirements that a human sign off on certain medical decisions.
The case against is that the mechanism is protectionism with better branding, and the objections are not new. When Gates proposed a robot tax in 2017, Larry Summers wrote in The Washington Post that Gates had gone "seriously astray," arguing that "Mr Gates' robot tax risks essentially being protectionism against progress" and that singling out robots made no logical sense when countless other innovations also destroy jobs. NYU Stern's Robert Seamans made the productivity version of the argument: taxes on automation "would dissuade firms from investing in robots, which would lower economic growth," and where robots complement labor, would mean less hiring, not more.
Then there is enforcement. Which jobs qualify, who decides, what stops a company from restructuring a reserved role into an unreserved one, and what happens when a country that reserves nothing competes against one that reserves 40 percent — these are load-bearing questions Gates does not answer. He acknowledges as much, writing that they "will need to be worked out in public." A blanket reservation could also freeze people into low-productivity work while wages stagnate around them.
What to Watch
Gates has not taken the proposals to the Trump administration, responding dryly when asked: "Well, you could tell me who at the White House I should be talking to about this." Federal posture runs the other direction — a June executive order asks AI companies to submit frontier models for government testing up to 30 days before release, but bars the program from becoming a licensing regime. The White House finalized that voluntary framework in early August. Gates called it "empty."
Watch whether any of this migrates into legislative text. The likelier near-term venues are state houses and sectoral regulators rather than Congress: staffing mandates in health care, human-review requirements in hiring and benefits adjudication, rules on AI in classrooms. Watch too whether the token tax finds a sponsor. Nine years after the first robot tax pitch, no major economy has enacted one — and the tax base Gates now wants to touch is the fastest-growing revenue line in American technology.
“In a very extreme form of it, I could imagine 40% of jobs could initially be reserved for humans. But that's as high as I can get.”— Bill Gates, Co-founder, Microsoft