Bill Gates sees artificial intelligence as both a source of potential progress and a force that could unsettle society for years. In an interview with The Verge, the Microsoft cofounder said the pace of change is outstripping the response, leaving too little preparation for harms even as beneficial uses may arrive slowly.
AI’s capabilities are advancing across different areas
Gates points to recent progress in code writing, finding cybersecurity flaws and reasoning about biology. He also cites AI’s ability to form relationships with children and other people, as well as its potential to disrupt the job market.
These capabilities make the issue broader than any single product or company, in his view. AI includes open-source and proprietary models, and the potential consequences depend in part on how those models are built, distributed and used.
Gates says society should prepare for these effects rather than treat them as remote possibilities. The same underlying technology can support useful applications and create opportunities for misuse, so questions about access and safeguards sit alongside questions about innovation.
Safeguards matter, especially in the near term
In the interview, Gates distinguishes between models that retain safeguards and models whose protections can be removed. He sees the latter as a more immediate danger, particularly where someone could use a model’s capabilities to support cyberattacks or bioterrorism.
Models with safeguards, monitoring and efforts to detect harmful use present a different level of near-term risk, he argues. But safeguards do not settle the longer-term control question: Gates says that concern applies to any state-of-the-art model.
That distinction complicates the idea that users can address the danger simply by choosing a company they trust. A model’s capabilities can matter beyond the preferences of ordinary customers, especially when someone intent on causing harm can seek out a system without effective protections.
Preparation needs to keep pace with the technology
Gates’s concern is not that AI can only cause harm. He agrees that AI can be beneficial if it is shaped in the right way, but he believes the current response is inadequate to the speed of development. Without better management, harmful outcomes may not be limited, while the benefits could take longer to reach people.
That puts both technical and policy questions on the table. Developers can build safeguards and monitor activity, while governments and society have to decide how to respond to risks that cross company boundaries. The interview raises a difficult enforcement problem: how can controls apply across both proprietary systems and open-source models?
Gates’s view, as presented in the conversation, is that the challenge cannot be reduced to trusting individual company leaders. The broader field is advancing, and efforts to manage its effects have to account for models and uses beyond any one organization.
A prolonged period of social change
Gates expects roughly two decades of AI disruption. He warns that the transition could be painful enough to divide society between people comfortable living near the technological frontier and those who reject it.
That possibility connects the technical debate to questions about work and generational attitudes. The interview also covers labor concerns and what might make people hopeful about the future, alongside a younger generation’s nihilism toward technological progress.
The central challenge in Gates’s argument is to make room for useful AI while preparing for disruption and misuse. That means treating safeguards, oversight and public readiness as part of the technology’s development, rather than waiting for its consequences to become impossible to ignore.