Bill Gates says artificial intelligence is about to change everything, and society is not ready. In a 6,000-word essay, the Microsoft co-founder warns that AI could destabilize the job market and put dangerous tools in the wrong hands. He argues that governments need a plan before the technology races further ahead. The bluntest version of his argument: AI is cheaper than workers, and the rules of the economy are currently rigged in its favor.

Gates’s essay begins with familiar alarms about AI’s unprecedented disruption. He writes that the technology could be as transformative as the PC or the internet, but unlike those earlier shifts, AI could replace not just manual labor but also cognitive work. Jobs in customer service, content production, legal research, and software development are already being automated. The pace, he says, is faster than any previous technological revolution.

Key facts at a glance

• Bill Gates published a 6,000-word essay on AI risks and policy responses.
• He proposes a category of “Human Reserved” jobs to protect work for people.
• He also calls for a tax on AI tokens and robots to reverse the current incentive to replace human workers.
• Gates argues the existing tax system rewards automation because employers can write off AI costs but must pay payroll taxes for humans.
• An AI token tax could fund retraining programs for displaced workers.

For all the grand warnings, Gates’s most provocative ideas come in the second half of the essay. The first is a notion he calls “Human Reserved” jobs. These would be categories of work reserved exclusively for biological humans, including roles such as caretakers, teachers, and social workers. The intention is to preserve meaningful employment and human connection in an increasingly automated world. Gates calls this a way to ensure that people still have a place in the economy, even if machines can do most productive work.

But the proposal raises immediate questions. Who would decide which jobs qualify as Human Reserved? How would those rules be enforced? What happens when employers try to automate around the designation? The very phrase “Human Reserved” is also a little unsettling. It suggests a future where humans are permitted to hold certain jobs only by explicit legal protection, rather than because they are the best person for the job. Critics might see it as a dystopian label, even if the intent is compassionate.

Gates is aware of these problems. He acknowledges that the idea needs more detail and that it may sound strange. Still, he clearly believes that relying on the free market alone will lead to painful outcomes, because the market does not naturally value human dignity. Without some intervention, companies will simply choose the cheapest option, and AI is already becoming the cheapest option for a growing number of tasks.

That brings Gates to his more concrete suggestion: an AI token tax. The idea is simple to describe but difficult to implement. Every time a business uses an AI model and processes tokens — the small chunks of text that language models read and generate — it would pay a tax. The same logic could apply to robots and other automation tools. The goal is to make employers think twice before swapping a human for a machine.

Gates points out that the current tax code is backwards. Employers must pay payroll taxes, unemployment insurance, and benefits for human workers. In contrast, spending on AI software and robots is often treated as a business expense, which can be written off and reduce taxable income. That gives companies a financial incentive to automate, even when a human worker might do a better job. A tax on AI tokens would flip that incentive. It would make automation more expensive and create new revenue for governments.

That revenue could be used to fund job retraining programs, wage subsidies, or even universal basic income experiments. As automation spreads, tens of millions of workers could be displaced across industries. The tax could soften the blow by supporting people through transitions to new careers or by providing a basic level of security while they search for work.

Big business would almost certainly hate an AI token tax. Technology companies that sell AI tools would see a new tax as a threat to their growth. Their customers would pay more, which would slow adoption. But there is also a large group of workers who are anxious about AI taking their jobs. Polls in recent years show that many employees worry about automation, not because they dislike technology, but because they doubt their employers will protect their livelihoods. For those workers, a tax on AI could be genuinely popular.

There is precedent for this line of thinking. Gates himself floated a robot tax back in 2017, in an interview with the website Quartz. His comments generated a fierce debate. Some economists said a robot tax would slow innovation and discourage productivity gains. Others said it was necessary to avoid a dystopian future where wealth flows only to the owners of capital. A few politicians and policy groups have proposed similar measures, but no major government has enacted one. The idea remains theoretical.

The AI token tax, however, is a new twist. Tokens are not the same as physical robots. They are the basic units of computation in large language models like ChatGPT, Claude, and Gemini. Every query a user sends to an AI chatbot is broken down into tokens, and the model generates a response one token at a time. If each token carried a small tax, a company running thousands or millions of AI interactions would quickly rack up a significant bill.

But implementation would be messy. Not all tokens are priced equally. A cheap model might charge a fraction of a cent per token, while a premium model could cost many times more. Would the tax be a flat amount per token, or would it vary by model? How would tax authorities keep pace with new model releases and pricing changes? Would certain uses be exempt, such as academic research or small businesses? And wouldn’t businesses find loopholes by moving their AI workloads to countries with lighter tax rules? These are not trivial questions.

There is also a deeper puzzle: how to measure the value being created by AI. A token used to generate a legal contract is not the same as a token used to generate a casual email. A flat tax would treat all tokens equally, which is unfair. A value-based tax would be incredibly complex. And any tax on inputs, rather than profits, can distort behavior in unexpected ways. Companies might use less efficient models simply to avoid the tax, or they might bundle AI costs into other services to hide them.

Gates does not pretend to have all the answers. He says the proposals are meant to start a conversation, not to be the final word. His broader point is that governments cannot remain passive while AI reshapes the economy. It would be irresponsible to assume that the market will produce fair outcomes on its own. The laissez-faire approach may work for new apps and gadgets, but it is dangerous when applied to something as powerful as artificial intelligence.

The historical record offers some reassurance. Past waves of automation, from the steam engine to the computer, ultimately created more jobs than they destroyed. But they also caused painful transitions. Workers who lost livelihoods during the Industrial Revolution did not simply walk into new careers overnight. Governments responded with public education, labor laws, and social safety nets. Gates argues that AI demands a similar response, and it needs to happen sooner rather than later.

The difference this time is that AI is not only automating physical labor but also reasoning, writing, drawing, and even coding. White-collar workers who once felt protected by their education now face disruption. The pace of change is far faster than earlier technological shifts. A new AI model can be adopted by millions of users within weeks, not decades. Policymakers do not have the luxury of waiting to see how the dust settles.

Gates’s “Human Reserved” idea may be dismissed as unrealistic, but it does raise a legitimate concern. Human beings want more than money. They want purpose, social status, and a sense of contributing to their communities. If AI eliminates the need for human labor, society will have to find new ways to provide those things. A protected class of jobs, however awkwardly defined, is one possible answer. The token tax is another. Both are attempts to deal with a future that many people are not prepared to face.

In the end, the most important takeaway from Gates’s essay may not be the specifics of any single proposal. It is the recognition that AI is about to make the old rules obsolete. The tax code, the education system, and the social safety net were all designed for a world where human labor was the dominant source of value. As AI becomes cheaper and more capable, those structures will need to change. Gates is using his influence to push for that change before the crisis arrives. Whether policymakers listen is a different question. The next few years will determine whether these ideas become the foundation of a new economic model or just a footnote in the history of AI.

Source: PCWorld News