Bill Gates is no longer talking about artificial intelligence as simply the next great productivity tool. He is warning that the technology could eliminate entire categories of work, strengthen cybercriminals and create biological dangers — and he wants governments to reserve some jobs exclusively for people.
In a new essay and an interview published Wednesday, the Microsoft co-founder called for national safety systems and a new international organization to oversee the most powerful AI models. Gates also said he hopes to discuss the proposal directly with Chinese President Xi Jinping during a tentatively planned trip to China in November, according to Reuters.
No meeting has been confirmed by Xi or China’s government. Gates’ staff told Reuters they are working to arrange one.
What Gates wants governments to monitor
Gates argues that the most capable AI systems should face coordinated testing before developers release them widely. One concern is whether an advanced model could help users design dangerous molecules or plan a biological attack. Another is whether it could autonomously break into computer systems or make cybercrime far more accessible.
He believes the United States should take the first step by defining when a model becomes dangerous enough to restrict. If Washington established a credible standard, Gates told Reuters, he thinks China and other governments might participate.
His proposed international body would borrow elements from organizations that oversee nuclear materials, aviation safety and protection of the ozone layer. The comparison does not mean AI would be regulated exactly like an airplane or nuclear facility. It reflects Gates’ view that no country can manage a globally distributed technology on its own.
The surprising proposal: jobs reserved for humans
Gates’ warning extends beyond national security. He is also concerned that AI could replace human labor faster than previous technologies created new occupations.
To soften that disruption, he proposes designating certain roles as “human reserved,” according to reports by The Guardian and the Financial Times. The idea resembles a protected reserve: even if software could technically perform a task, society might decide that some responsibilities should remain with people.
Caregiving is an obvious example because empathy, trust and human presence are central to the work. The larger question is whether governments should protect additional professions — and who would decide which ones qualify.
Gates also raised the possibility of changing tax systems so that companies benefiting from AI contribute more toward income support, retraining and other protections for displaced workers. Those ideas remain proposals, not adopted policies.
Why his tone has changed
Gates has long emphasized AI’s potential to expand access to health care, education and scientific research. He still argues that those benefits could be transformative, particularly in countries with shortages of doctors and teachers.
But his latest intervention puts greater weight on the speed of the transition. Fortune reported that Gates is particularly worried about disappearing entry-level opportunities, reduced critical thinking among students, emotionally manipulative chatbots and increasingly capable criminals.
Those concerns arrive while AI companies are releasing stronger models at a rapid pace and governments are still debating basic standards for testing, transparency and responsibility when systems cause harm.
A global plan would face major obstacles
Creating an international AI regulator would require rivals to share information about some of their most strategically valuable technology. The United States and China are simultaneously competing over chips, data centers, cloud computing and advanced models, making sustained cooperation difficult.
There is also no global agreement on what counts as an unacceptable AI capability. A model useful for drug discovery might also be misused to design harmful substances. Security testing could reduce danger, but overly broad restrictions could slow medical or scientific work.
Gates has offered a framework, not a finished rulebook. His most consequential point is that leaving every decision to individual companies may be inadequate when the possible effects include biological threats, mass job displacement and attacks that cross national borders.
The debate is therefore shifting from whether AI will transform society to who will set the limits — and whether governments can act before the most disruptive capabilities arrive.
