Bill Gates: 'There Is No Plan' — Inside the Essay That Reframed the AI Debate

Two days after it was published, Bill Gates’ essay “The turbulent AI era is here. The choices we make now are critical.” is still the most debated document in AI policy. Posted August 26 on Gates Notes and syndicated to LinkedIn — where it has drawn nearly 8,700 reactions and more than 1,700 comments — the piece does something unusual for a tech billionaire: it argues that the industry that made his fortune is systematically understating what is coming, and that governments have, in his words, “no plan to ease the entry into the AI era.”

The essay arrives at a moment of visibly shifting public sentiment. In a July POLITICO poll, 41 percent of Americans said they would oppose a data center being built near their home, against just 24 percent who would support it. Midterm candidates — and a handful of 2028 presidential hopefuls — have begun campaigning on making data center construction harder. Gates’ intervention gives that unease a framework, and a shopping list of fixes.

“This time really is different”

Gates spends the first half of the essay dismantling the comfortable analogy that AI is just another technology transition, like the shift from agriculture to office work. That transition, he notes, played out “over several generations and created new jobs where human cognition was required.” AI, by contrast, “can substitute for human cognition.” Because it sees, listens, speaks, and reasons — and will eventually do physical work “just as smoothly as any human” — it will not confine itself to one sector. Law, customer service, medicine, software, and manufacturing will all be hit, and hit “over the course of a decade rather than a few generations.”

He also acknowledges why so many commentators underestimate the impact: models still make mistakes, and it is hard to take seriously a technology that until recently could not count the R’s in “strawberry.” But the reliability problem “is being fixed quickly, as researchers create models that can check their own work,” and adoption requires none of the plumbing that slowed the PC era — AI runs on devices people already own and speaks natural language. “We don’t have to adapt to it because it can adapt to us.”

One passage stands out for its bluntness: “If someone had a credible plan for slowing down AI advances globally, I would likely support it. However, I don’t think that’s going to happen. The geopolitical and economic incentives are pushing too hard to go full speed ahead.” That single sentence split the difference between AI’s boosters and its doomers — and generated most of the headlines.

Three risks, quantified by comparison

Gates structures the danger around three risks:

1. Many jobs will disappear forever. His reference point is the Great Depression, when U.S. unemployment hit roughly 25 percent in 1933 and stayed in double digits for much of the following decade. AI may not reach that level, he concedes — but unlike a cyclical downturn, “its impact will not go away with an economic cycle.” The jobs at most risk are entry- and mid-level, and the new jobs being created “will mostly require skills that take many years to learn.” He cites evidence that after the widespread adoption of generative AI, employment fell significantly among young workers in vulnerable jobs “but not among their older colleagues” — and predicts the pattern will spread from sales, customer support, software engineering, and paralegal work into loan assessment, data analysis, and even patient triage. Blue-collar work is not exempt: he expects “smart” robots to compete on physical tasks in construction and hospitality “by the end of the decade,” and warns of a vicious cycle where one adopter’s cost savings force every competitor to follow.

2. AI will empower bad actors. The same model that finds a software flaw so it can be fixed can help a criminal exploit it. Gates says the smartest cybersecurity experts he knows “are scared about the next few years,” because attackers are gaining capabilities faster than defenders can patch weaknesses. The target list is infrastructure: hospitals, financial institutions, water systems, power grids, benefits systems. And the bioterrorism argument mirrors the cybersecurity one — the same capabilities that accelerate drug discovery make it easier to design a deadly new disease. “The positive capabilities are hard to separate from the dangerous ones.”

3. AI could stunt child development. Gates, who says he worked hard as a teenager in Seattle to develop his social skills, doubts he would have put in that work with an AI companion available — companions “don’t push you outside your comfort zone.” He cites a Stanford/Carnegie Mellon study of over 1,100 AI companion users, which found that those with smaller social networks were most likely to turn to chatbots, and that heavier, more emotionally personal use correlated with feeling worse. He also flags preliminary research linking heavier AI use to weaker critical thinking — “the worst possible time for humans to lose” that skill.

Three proposals — including a robot tax

The essay’s second half is what policy circles are actually arguing about. Gates offers three ideas:

Build a new governance system, domestic and international. His highest-priority ask is “a monumental task: creating a domestic and international framework for dealing with AI.” He compares the required reorganization to the post-9/11 creation of the national security apparatus — “AI will require much, much more” — because it cuts across employment, education, taxation, energy, elections, public health, finance, and law enforcement simultaneously. Internationally, he points to existing models: nuclear weapons inspections, international aviation regulation, and ozone-layer agreements. A new global AI institution “will need elements of all three and more.” Notably, he writes that some U.S.–China cooperation “will be required” — and that countries hosting the leading developers should start setting shared norms “before competitive pressure makes it harder to cooperate.”

Set aside some jobs for humans. In the essay’s most original section, Gates proposes a concept he calls “Human Reserved” — jobs deliberately protected from automation, “places where we could put buildings and roads, but we choose not to because the loss would be too great.” The idea is partly economic (a 55-year-old construction worker cannot simply retrain into elder care) and partly moral: “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.” The inspiration is personal — his father’s Alzheimer’s caregivers in his final years, care that was “irreplaceably human.” He is candid that the concept raises unanswered questions: Who decides what gets reserved? How do you stop companies from cheating? What happens to trade when one country allows robots in a sector and another doesn’t?

Rebalance how we tax labor and capital. Gates renews his call — first made years ago, to general dismissal — for taxing AI tokens and robots. The logic: an employer pays payroll taxes on a human worker’s earnings but can typically write off a robot as an immediate business expense, so “the tax system nudges you toward replacing people with machines.” A tax would slow that rush slightly and fund retraining plus a stronger safety net, targeted so it doesn’t slow beneficial uses like cheaper medicine and education. He concedes economists call it inefficient, and answers that “we’ll be able to afford a little inefficiency as the price for keeping people employed.”

The balance sheet

Gates is careful to pair the warnings with upside. When intelligence stops being the limiting factor in R&D, “smaller companies will be able to compete with organizations that have far larger research budgets.” He highlights Viz.ai, whose stroke-detection and care-coordination tools are used in nearly 2,000 U.S. hospitals; AI advice for low-income farmers that could soon outclass what even the richest farmers get today; streamlined government benefits applications; and AI tutors designed to preserve “productive struggle.” The Gates Foundation — 19 years into its final 20, with $200 billion left to spend — will use AI to accelerate HIV, TB, malaria, and malnutrition work, with OpenAI, Anthropic, Google, and Microsoft all partnering on foundation initiatives.

He also discloses his own position with unusual candor: he still holds financial ties to the tech industry, but all profits from his investments go to the foundation, and “readers will have to decide for themselves whether this clouds my view.” And he notes approvingly Pope Leo XIV’s encyclical “On Safeguarding the Human Person in the Time of Artificial Intelligence” as “a strong foundation for the work that needs to be done.”

Why it matters

The significance of this essay is less its novelty than its messenger and its timing. The person who wrote the software that defined the office-worker era is now saying that era’s work is structurally over — and that the disruption will arrive in years, not generations. POLITICO framed the stakes plainly: Gates would support slowing AI down if a credible global plan existed, but doesn’t believe one will. That leaves preparation as the only lever — and Gates’ message to leaders is that they have “a chance to act now, before unemployment rises sharply, communities are hurting, and public trust has eroded.”

Whether “Human Reserved” or a robot tax ever enters legislation is doubtful in the near term. But the essay has already shifted the Overton window of respectable AI policy conversation — away from “will jobs disappear?” and toward “what do we deliberately choose to keep?”