Bill Gates Wants a Face-to-Face With Xi as His AI Fears Deepen
Bill Gates AI warning
Bill Gates Fears AI Could Destroy Jobs and Empower Attackers — Now He Wants Global Rules
Bill Gates spent decades believing technology could solve almost any problem humanity placed in front of it. Artificial intelligence has pushed him into unfamiliar territory: he now appears increasingly worried that innovation may be moving faster than the institutions capable of controlling its consequences.
The Microsoft co-founder is seeking a meeting with Chinese President Xi Jinping later this year as he pushes an extraordinary idea into the centre of the global AI race. Gates believes the United States and China may still be able to cooperate on restricting the most dangerous artificial intelligence capabilities, even while the two countries compete intensely to dominate the technology itself.
Gates Wants Xi Jinping at the Table
Gates is looking to meet Xi during a future trip to China, tentatively expected later this year. His aim is not simply another discussion about technology investment or philanthropy but a conversation about whether the world’s leading powers can establish shared safeguards before AI capabilities become significantly harder to contain.
One of Gates’ central arguments is that Washington may have to move first. He believes Beijing could potentially accept restrictions on the release of particularly dangerous AI models if the United States established similar restrictions itself, creating the foundations for rules that extend beyond either country.
That would represent a striking break from the dominant narrative surrounding the US-China AI rivalry. Washington and Beijing are currently trying to secure advantages in advanced chips, computing power, data centres, robotics, industrial deployment and strategic technologies. Gates is effectively arguing that competition does not remove the need for cooperation when the potential consequences cross borders.
That distinction could become crucial. A country may have every reason to race ahead economically while still having no interest in allowing technology capable of enabling catastrophic attacks to become freely available to anyone able to download or reproduce it.
The Fear Is No Longer Science Fiction
Gates’ concern stretches far beyond chatbots producing inaccurate answers or students using AI to write essays. His focus increasingly includes cyberattacks, biological threats, economic disruption and the possibility that highly capable systems could be exploited by people with relatively little specialist knowledge.
Biology appears particularly important. Gates wants countries to monitor whether advanced systems become capable of designing molecules or materially assisting biological attacks, creating an early-warning mechanism around a capability that could carry consequences far beyond the technology industry.
The fear is not that an AI system inevitably decides to launch a biological attack itself. The more immediate concern is amplification: a sufficiently capable system could potentially place expertise that once required specialist laboratories, extensive training or large organisations into the hands of much smaller groups.
Gates has previously warned that AI could make a non-state biological attack more plausible, while simultaneously acknowledging the enormous potential for the same technology to accelerate drug discovery, healthcare and disease prevention. That creates the central problem running through much of the modern AI debate: the capabilities society most wants may be closely connected to capabilities it most fears.
It is a problem already visible in the collision between artificial intelligence and biological security. The question is no longer simply whether AI becomes more powerful. It is whether governments can distinguish beneficial capability from dangerous capability quickly enough to act.
Gates’ Position on AI Is Getting Harder
This matters partly because Gates has traditionally been one of the technology world’s most powerful optimists.
He has repeatedly argued that artificial intelligence could transform healthcare, education, agriculture and economic productivity. His foundation is actively putting that belief into practice, including partnerships designed to deploy AI in healthcare and development.
That optimism has not disappeared.
But it now comes with a much sharper warning.
Gates believes society is approaching a period in which artificial intelligence could reorganise work at extraordinary speed. Earlier technological revolutions replaced particular forms of labour while creating entirely new categories of employment. AI presents a different possibility because increasingly capable systems can potentially perform the cognitive tasks that might otherwise have become the replacement jobs.
Software development is already one area Gates believes is being disrupted. Customer support, administrative work and other knowledge-intensive jobs could follow, while improvements in robotics could eventually push the same transformation into warehouses, construction, hospitality and other physical industries.
The deeper issue is speed.
Societies can adapt to technological change. Workers retrain. Industries emerge. Education systems evolve. Tax systems change. New forms of employment appear.
They struggle far more when all of those adjustments are required simultaneously.
That is why the AI transformation of work is becoming as much a political question as an economic one.
Gates Is Considering Something Once Almost Unthinkable
One of Gates’ most provocative ideas is that societies may deliberately decide that some jobs should remain human.
If machines become capable of providing particular services more cheaply and perhaps even more consistently than people, pure economic logic would normally favour automation. Gates is questioning whether pure economic logic should determine everything.
Care is the obvious example.
A machine might eventually monitor a patient more continuously than a nurse, recall more medical information than a doctor and respond instantly at any hour. None of that necessarily means society would want human contact removed from medicine, elderly care or other deeply personal settings.
Gates has therefore floated the concept of preserving a substantial share of jobs for humans and has also discussed changes to taxation that could redistribute some of the gains from machine labour.
That takes the AI debate somewhere much bigger than technology regulation.
It raises a political question about what work is actually for.
If machines eventually become capable of producing enormous quantities of goods, services and knowledge with dramatically less human labour, the economic benefit could be enormous. But if ownership of those machines is concentrated while wages remain the primary mechanism through which ordinary people receive income, the same productivity revolution could create severe instability.
AI could make society vastly richer while leaving millions of people feeling economically unnecessary.
That contradiction may be one of the defining political problems of the next decade.
Why China Matters So Much
Any global agreement on advanced AI that excluded China would have an obvious weakness.
China possesses enormous computing resources, world-class researchers, major technology companies, a huge domestic market and a government determined to make artificial intelligence central to the country’s economic and strategic future. Beijing is pushing AI deeper into manufacturing, public services, robotics and industrial systems through a scale of national mobilisation few countries could reproduce.
The scale of China’s AI push means restrictions adopted solely in the West could simply alter where particular capabilities are developed rather than controlling them globally.
Gates therefore appears to be pursuing something closer to arms-control logic.
Nuclear powers compete fiercely, yet they have historically maintained inspection systems, treaties, communication mechanisms and agreed boundaries because some risks are too severe to leave entirely to competition.
Artificial intelligence is not nuclear technology, and any international AI regime would face different technical and political problems. Models can be copied. Software crosses borders easily. Computing capacity is distributed. Verification can be difficult.
But Gates’ underlying argument is becoming clear: AI may now be powerful enough that international rivalry alone is an inadequate governing system.
Washington Would Have to Trust Beijing — And Beijing Washington
That is where the plan becomes difficult.
The United States considers advanced artificial intelligence strategically important. China does too. Both governments have powerful incentives to fear that restrictions accepted by one side could be exploited by the other.
Any meaningful agreement would therefore need more than political promises.
There would have to be some method of establishing what capabilities were restricted, how compliance could be checked, when a model became dangerous enough to trigger controls and what happened when either side believed the other was cheating.
Even defining a dangerous model could prove contentious.
Is danger determined by computing power? Performance in cybersecurity tests? Ability to assist biological research? Autonomous behaviour? The number of users? Whether model weights are publicly released?
AI moves rapidly enough that yesterday’s sophisticated benchmark can become tomorrow’s ordinary consumer capability.
Gates is not claiming to have solved those questions. What is significant is that one of the most influential figures in the history of computing increasingly believes they require international political machinery rather than voluntary promises from technology companies alone.
The Contradiction at the Heart of Gates’ Warning
There is an unavoidable contradiction in Gates’ position, but it is also what makes his warning important.
He remains deeply invested in the promise of artificial intelligence.
The Gates Foundation intends to spend enormous sums before its planned closure in 2045, with AI playing an expanding role in projects ranging from health systems to education and agriculture. Its partnerships with major AI developers are based on the belief that machine intelligence can extend expertise into places suffering from chronic shortages of doctors, teachers and other skilled workers.
AI could therefore save lives at the same time as creating new biological risks.
It could expand access to expertise while destroying traditional professional jobs.
It could make governments more efficient while making surveillance dramatically more powerful.
It could increase productivity while concentrating wealth.
It could become one of history’s great engines of equality or deepen existing inequality on an extraordinary scale.
That is why Gates’ position is more significant than a simple conversion from AI optimist to AI pessimist.
He is becoming an AI conditionalist.
The technology can still deliver extraordinary benefits, in his view, but those benefits can no longer be assumed to arrive automatically.
The Gates-Xi Meeting Could Matter Far Beyond One Conversation
Whether Gates actually secures the meeting he wants remains important. So does whether Xi has any interest in the proposals.
Beijing has not publicly committed to Gates’ vision of international restrictions on dangerous model releases. Nor is there evidence yet of a US-China agreement waiting to be activated.
But Gates has something relatively unusual in the deteriorating relationship between the two countries: personal access.
He has met Xi before and has maintained longstanding involvement in China through technology and philanthropy. That does not make him a government negotiator, and any binding agreement would ultimately have to involve states rather than a private citizen.
It does, however, give him the ability to put ideas directly in front of one of the few people capable of materially changing China’s approach.
That may matter if AI safety eventually reaches the point where Washington and Beijing decide that neither can protect itself without some cooperation from the other.
The world spent the first years of the AI boom asking who would win.
Gates is increasingly asking a different question.
What happens if everybody wins the race to build enormously powerful AI before anybody agrees on the rules for controlling it?
