Strikingly, one of the most insightful analyses of the promise and peril of AGI came from OpenAI CEO Sam Altman. In a blog post originally published in 2023 and updated in 2025, Altman displayed a measure of philosophical skepticism uncommon among tech optimists. “We want AGI to empower humanity to maximally flourish in the universe,” he wrote, while recognizing that doing so would require “successfully navigating massive risks.”
To be sure, some skepticism is warranted when assessing the potential benefits of technological advances. Since the Enlightenment, humanity has increasingly rejected superstition and dogma in favor of the skeptical spirit of scientific inquiry, and, over time, nurtured the belief that scientific progress is inherently beneficial. While individual discoveries like the atomic bomb can be destructive, the thinking goes, science as a whole, which gives us a deeper understanding of nature, is as an unqualified force for good and, by extension, greater human welfare.
What is often overlooked, however, is that unquestioning faith in science can itself become a form of dogma. In a 2024 paper, Jörgen Weibull and I used game theory to explore this dynamic. When humans interact under conditions of uncertainty, where what is rational for one person depends on what is rational for others, a new scientific discovery can uncover situations that resemble the classic prisoner’s dilemma, in which individually rational behavior leads to outcomes that leave everyone worse off. The rise of AGI could prove to be one such event.
Altman’s skepticism is therefore well placed. As he notes, AGI “has the potential to give everyone incredible new capabilities,” but it could also lead to “drastic accidents and societal disruptions.” The challenge, then, is to minimize the risk of a “knowledge curse.” Given that the future remains uncertain, the best we can do is make informed guesses and adopt prudent preventive measures. Altman is right to argue that these challenges cannot be left solely to the private sector. Addressing them will require some degree of resource redistribution and concerted international cooperation.
Perhaps the most serious likely consequence of AGI is a decline in the demand for labor. Crucially, this does not mean people will have nothing to do; indeed, they may enjoy unprecedented amounts of leisure. But if labor is defined as it is in economics textbooks—as any activity undertaken for compensation—an AGI-powered world in which machines outperform humans in virtually every domain could render it obsolete.
Since the vast majority of adults depend on labor income to survive, such a transformation could deprive billions of people of their livelihoods. For this reason, Altman has advocated a universal basic income (UBI), which would guarantee every person a minimum standard of living.
Yet even that may not be sufficient. Beyond its impact on jobs and incomes, AGI could pave the way for an era of global authoritarianism. If control of these systems ends up highly concentrated, a tiny elite of mega-billionaires could wield unprecedented power over all of humanity. As Altman argues, preventing such an outcome requires the “benefits of, access to, and governance of AGI to be widely and fairly shared.”
As with UBI, broad access to AGI is only part of the solution. It is equally important to prevent extreme concentrations of wealth and power that could push societies past the point at which democracy gives way to authoritarianism. To this end, what may be needed is a universal basic share: a system that limits how much of the world’s wealth any individual or small group can hold.
This matters not only for the poor but also for many affluent people. In a world where a few thousand individuals control the levers of AGI, even today’s millionaires could find themselves among tomorrow’s underclasses.
