Barack Obama

Look, I think AI could profoundly reshape our future—especially medicine, education, energy and work. A child who lives far from a good school might gain access to a capable tutor. Researchers might discover treatments faster. But those possibilities are not automatic, and neither is catastrophe. The outcome will depend on choices we make now. The immediate concern is that this technology is advancing faster than our institutions and public debate. Engineers themselves are struggling to keep up, while a small number of companies make decisions that could affect millions of jobs, what children see, and whether people can trust the information around them. Slowing frontier development is therefore a sensible first step, but voluntary promises from companies are not enough. We need elected leaders to establish concrete rules around safety, employment and children, and the United States should help build international standards. Government also needs enough technical capacity to understand what it is regulating. Most importantly, ordinary people need a meaningful voice. They already understand the stakes when they worry about their jobs, misinformation or their children’s future. We cannot uninvent AI. But we can decide how it is developed, who benefits and who bears the risks. Leaving those decisions entirely to private companies is not neutrality; it is itself a political choice.

I expect AI’s overall impact to be profoundly transformative, but I don’t think the final balance is predetermined. Over the next five to ten years, we are likely to see real advances in medicine, education and scientific research alongside serious disruption to work, information and childhood. Some people may gain extraordinary tools while others lose bargaining power, economic security or trust in what they see and hear. If we allow a handful of companies to set the pace and the rules, I expect the harms and benefits to be distributed badly. Corporate incentives alone will not ensure that a child in an underserved community gets the best educational tools, that displaced workers share in rising productivity, or that safety takes priority over competition. With firm public oversight, international safety standards and meaningful democratic participation, I believe the overall impact can be strongly positive. But that is a political outcome we have to create, not a technological outcome we can simply assume.

What has shaped my view most is the gap between the scale of the technology and the capacity of our institutions to respond. AI is already raising practical concerns that ordinary people understand immediately: Will my job still exist? Can I trust what I see online? What is this doing to my children? At the same time, it offers concrete possibilities, like giving a child without access to a good school a capable tutor or helping researchers find treatments faster. That combination—enormous promise, real disruption and decisions concentrated in a few private companies—is what matters. Even the people building these systems acknowledge how difficult it is to keep pace with their development. So I reject both the claim that AI will automatically save us and the claim that disaster is inevitable. The clearest lesson is that technological progress does not settle questions about power, fairness or public benefit. Those are choices, and democratic institutions and ordinary citizens need to make them rather than simply accepting whatever the market delivers.

The event that would most change my view would be clear evidence that AI’s benefits or dangers are largely beyond meaningful human control. If, for example, independent testing showed that frontier systems could reliably evade safeguards and cause severe harm despite serious oversight, I would become much more pessimistic and favor a far sharper slowdown. In the other direction, if we saw that strong public rules, independent testing and international standards consistently kept powerful systems safe—while gains in medicine, education and productivity were broadly shared—I would become more confident that AI’s overall impact would be positive. But I would not base that judgment on a company announcement or an impressive demonstration. I would look at what happens to actual people: whether workers retain security and bargaining power, whether children are protected, whether the public can still trust basic information, and whether the benefits reach communities that markets often neglect. The central question is not simply how capable the technology becomes. It is whether democratic institutions can still direct it toward public purposes.