Bill Gates just put two ideas on the table that could reshape how governments respond to AI’s effect on jobs: a “robot tax” and a category of work he calls “Human Reserved.” He laid them out in a long essay on his Gates Notes site, and as TechCrunch AI reports, most of it tracks with the Responsible AI camp. What stands out are the two proposals that go further than the usual talking points.
Gates isn’t a doomer here. He’s excited about what AI can do for science and healthcare, and he backs the spirit of “Pacing the Frontier,” the open letter from AI employees pushing for a slowdown. He’s just skeptical a voluntary slowdown holds. So he’s reaching for policy instead.
The robot tax, explained
The logic is about tax incentives, not machines being evil. Right now the system quietly pushes employers toward automation. Hire a person, you pay payroll taxes on their earnings. Buy a robot, you usually write it off as a business expense right away.
One path costs you every payday. The other gives you a deduction. Gates argues that tilt nudges companies to replace people with machines, and a robot tax would do two things:
- Slow the rush away from human labor, at least a little.
- Raise money for retraining and a stronger safety net.
It’s a small thumb on the scale, not a ban. That framing matters, because it’s the kind of thing a treasury department can actually write into code.
“Human Reserved” jobs
The second idea is blunter: set aside certain tasks and bar AI from doing them. Gates gives two reasons a job might land in this bucket.
The first is economic. Some workers can’t just pivot. As he puts it, “You can’t tell a 55-year-old who has worked in construction their whole career that they need to go work at an elder care facility and expect them to find it fulfilling.”
The second is human dignity. Picture a robot telling you that you have an incurable disease. There’s no technical reason it couldn’t. Gates argues it shouldn’t. Some moments should stay person to person, full stop.
He also suggests the list would evolve, reserving some roles now and phasing AI in slowly over years or decades, with a commitment to keep some jobs human on purpose.
Why this lands now
Here’s the part worth sitting with. TechCrunch AI points out that both ideas would put a serious dent in the profits of the major labs. Slower automation and protected job categories mean fewer seats for AI to fill, which is exactly the growth story labs are selling to investors.
That may be why you haven’t heard these proposals much until now. When one of Microsoft’s co-founders, an AI optimist, floats measures that cut against the labs’ revenue, it changes who’s allowed to say them out loud. This is significant because it moves the labor debate from activists and unions to the center of the industry.
It also arrives as the automation timeline compresses. Coding assistants, customer support agents, and back-office tools are already shipping. The gap between “AI could do this job” and “AI is doing this job” is closing faster than most policy moves.
The open questions
Neither idea is close to ready. The essay leaves the hard parts blank:
- Who sets the rules? A federal agency, states, some new body?
- What exactly counts as a robot for tax purposes? Software agents blur that line fast.
- How do you define a Human Reserved job without freezing whole industries in place?
Those aren’t small details. They’re the whole fight.
What to do about it
If you run a business or build with AI, treat this as an early signal, not noise:
- Expect automation incentives to become a political target. The write-it-off-immediately math on AI tooling may not last.
- Watch for “human in the loop” to shift from a best practice to a possible legal line, especially in healthcare, hiring, and anything touching vulnerable people.
- Build retraining and redeployment into your automation plans now, before regulators build it for you.
Gates is describing the next three years of the AI labor debate before it fully arrives. Whether these specific tools get built matters less than the fact that the conversation is moving from whether to slow AI to how. You can read his full essay at the original source.