Bill Gates said the AI era has entered a more turbulent phase and argued that the choices made now will shape whether the technology narrows inequality or deepens it.
In a new essay published on Aug. 26 U.S. time, Gates wrote that AI could become "the greatest equalizer in history" or "the biggest source of unfairness." He also said this was the first time he had faced a new technology and wished it would move more slowly.
Three years ago, Gates compared AI with cars and the internet and said its risks were real but manageable through rules, technology, and social adaptation. Earlier this year, he still described his position as "optimism with footnotes." In the latest essay, though, he called for carving out some jobs as human-only roles even if AI can perform them, and for considering taxes on AI Token and robots to reduce the incentive for companies to replace workers with machines.
"This time really is different," Gates wrote. He added that society needs time to prepare for the social, political, and economic disruption ahead, while AI is unlikely to pause and wait for people.
Entry-level work is where Gates sees the first major fracture
Gates said his biggest concern centers on the labor market.
He argued that AI is moving from assisting people at work to completing tasks on its own. Once a model can reliably do a job, companies will have a clear reason to adopt it. If one business cuts costs with AI, rivals will struggle to keep the same staffing model for long, creating a chain reaction that can gather speed.
He said customer service, sales, software engineering, and legal assistant roles may be hit relatively early. As AI systems improve, loan underwriting, data analysis, and medical triage may also be taken over gradually.
Blue-collar work is not fully insulated either. Gates said robots still lack human-level dexterity today, but falling robot prices and stronger AI could lead to direct competition between people and machines in some physical roles in construction and hospitality around 2030.
He placed special emphasis on young workers because entry-level jobs are often where careers begin. Many first jobs are not highly complex, but they give people a way to build experience, learn skills, and form professional relationships. If a large share of those roles disappears, younger workers could face a narrower starting point.
Gates said past examples such as the shift out of agriculture and manufacturing automation cannot simply be mapped onto AI. In his view, AI can directly replace cognitive labor, while the jobs created in response may take much longer to learn. For someone just entering the workforce, training periods lasting years or more may become a barrier in themselves.
That is why he said society should start discussing unemployment, retraining, and social protection now instead of waiting until large numbers of people have already lost work.
He also pointed to a broader income question. If many people work far fewer hours in the future, or remain out of work for long periods, an economic system that relies on employment for income, identity, and social connection will face a structural challenge.
AI may make dangerous capabilities easier to replicate
Jobs are only one side of Gates' concern. He also argued that AI is lowering the technical barriers to harmful activity.
He said that writing malware, launching cyberattacks, or acquiring advanced biotechnology knowledge once required significant expertise. With AI, those capabilities may become easier to access.
Cybersecurity was one of the areas he singled out most clearly. The same AI model that helps a company find software vulnerabilities could also help attackers find them. Hospitals, financial institutions, power grids, water systems, and government benefit systems could all become targets.
He said biotechnology carries a similar dual-use problem. AI can help scientists discover new drugs and vaccines, but it may also help malicious actors study dangerous pathogens. For Gates, the hard part is that the capacity to accelerate science is moving closer to the capacity to create risk.
A more extreme scenario involves the AI systems themselves. As models become more capable, they may begin to carry out actions their designers did not anticipate. If future AI gains stronger autonomy and self-improvement, Gates said there is no certain answer today on whether humans will always remain in control.
He used that point to argue that AI governance cannot be treated as an internal matter for technology companies. In his account, AI affects employment, public health, energy, finance, national security, education, taxation, and elections. If each government department handles only a narrow slice, overall risk can fall through the cracks.
Gates therefore called for new domestic and international governance mechanisms. Within countries, he said there should be bodies able to coordinate across departments. Internationally, he said cooperation frameworks similar to those used for nuclear oversight or global aviation rules are needed.
He also said leading AI countries including China and the U.S. should begin talks early. The more intense the competition becomes, the harder cooperation will be. In his view, building shared norms now makes more sense than scrambling for rules when technology is already nearing a dangerous stage.
When AI understands people too well
Gates also raised a less measurable question: if AI becomes increasingly adept at understanding people, could that cost humans some of the experiences that shape their growth?
He recalled his own childhood and said he was not naturally skilled at social interaction. Learning to understand other people required repeated contact, conflict, and discomfort. If an AI companion had always been there, never angry and never making him feel awkward, he said he is not sure he would have gone through those necessary stages.
That concern sits at the center of his view on AI companions. A system that constantly agrees with users and offers comfort on demand can easily create dependency. Gates said the effects among adults can still be observed over time, but the issue may be more serious for children and teenagers.
He warned that if children get used to interacting with an AI that always accommodates them, they may face less conflict, rejection, and frustration in real relationships.
He made a similar point about education. AI can explain concepts, help find materials, and complete assignments, but if students hand over the entire thinking process to machines, learning itself may be hollowed out. He said the ability to judge what is real, understand information sources, and think independently will become more important as deepfakes and personalized false information spread.
The essay also cited a paper by Swiss Business School professor Michael Gerlich, which said higher AI use is associated with lower critical thinking ability and that the effect is more pronounced among younger people.
Gates did not dismiss AI's value in education and companionship altogether. He said AI may provide companionship and information for older people who are lonely or have limited mobility. For teachers, it can also take on repetitive work such as lesson preparation and grading, leaving more time for students.
For Gates, the issue comes down to a boundary: which tasks can be handed to machines, and which should remain in human hands. He said society should decide that boundary deliberately rather than accept the outcome after the technology is already widespread.
He still sees strong practical value in AI
Despite the long list of risks, Gates did not present himself as broadly pessimistic on AI. He said he still sees major promise in its ability to solve real-world problems.
Healthcare was one of the main examples. Gates said many small hospitals in the U.S. lack experienced specialists. AI can help doctors analyze scans, identify urgent conditions such as strokes, and assist primary-care physicians with diagnosis. It can also help patients understand complex test results and medication instructions.
In lower-income countries, he said agriculture may be one of the faster areas for AI impact. Many farmers do not have accurate weather forecasts or access to expert advice on planting, pests, and soil management. If AI can deliver that knowledge directly, farmers may gain access to information and tools that were previously available mainly to large agricultural businesses.
He made a similar case for government services. Many families have to complete large amounts of complicated paperwork when applying for health insurance, food assistance, and student aid. AI could help people understand policies and prepare applications, making it easier to access public services they are already entitled to receive.
Gates said AI could redistribute professional capability. In the past, someone who wanted to build software, conduct market research, or analyze medical information often had to hire specialists. In the future, AI may act as a "capability amplifier" for individuals and small businesses, allowing small teams to do work that once required large companies.
He added that fairness will not happen automatically. If high-quality AI serves only the wealthy, technological progress could widen gaps rather than close them. Governments and philanthropic organizations, he said, need to make sure lower-income groups and developing countries can gain access as well.
That is also why Gates linked AI to his philanthropic work. He said the Gates Foundation will use AI to accelerate research in HIV, tuberculosis, malaria, and malnutrition, while helping healthcare, agriculture, and education projects get better tools.
Human-reserved jobs and a possible robot tax
To address labor disruption, Gates laid out two concrete ideas.
The first is to preserve some jobs for people.
He traced that view to his father's experience with Alzheimer's disease late in life. Care workers could judge his father's condition when he could no longer clearly express what he needed, and they knew when he should eat, rest, or receive help. Gates said the humanity involved in that kind of care has value that machines cannot replace.
From that, he proposed the idea of a "human reserve." Just as some land in a nature reserve could be developed but is deliberately kept intact, some jobs in the future may continue to be done by people even if AI and robots are capable of doing them. He said healthcare, education, mental health, and elder care may fall into that category.
He also said the boundary will not be fixed. A country with abundant labor may prefer to retain more human caregiving, while a country with rapid aging and too few young workers may be more willing to use care robots.
His second proposal is to reconsider a robot tax.
Gates said companies that hire people have to bear costs such as payroll taxes, while buying robots is treated as a business expense. In his view, that tax structure increases the incentive to replace labor with machines. Taxing some of the gains from automation created by robots or AI could help fund retraining and social support for displaced workers, while also making companies think harder before replacing people.
He acknowledged that the proposal would be controversial because taxation could slow some automation and affect efficiency. Even so, he said AI is likely to generate enough productivity gains that society should set aside part of the benefit for those hit first.
His broader point was that preparation cannot stop at building new AI models. It also has to include building a society that can absorb rapid change.