Why Calls for AI Regulation Echo Historical Technological Shifts

AI companies are increasingly asking for greater regulation, following a long line of technologists and industrialists who have done the same when the pace of digital development threatens to outstrip governing guidelines. In a recent essay titled We Must Pace the Frontier, Dario Amodei, the chief executive of Anthropic, argues that AI companies should introduce independent safety evaluators, coordinate on common safety standards, and work toward international cooperation. He has also proposed giving governments greater authority over particularly dangerous AI deployments, moving beyond asking individual companies to adopt more safety policies to call for collective rules and external oversight.

At first glance, this might seem strange. Why would competing companies want governments to put constraints on their own industry? However, economic history shows how prior waves of technological innovation have often reached tipping points where booms and busts push for greater oversight. Industrial development consistently moves faster than legislation, and governments tend to catch up only after technology becomes deeply embedded in society. We saw this most recently with the internet and social media during the late 1990s and early 2000s, where companies and regulators argued over privacy and consumer protection as the industry relied heavily on self-regulation.

Today, AI is entering a similar phase, and Dario Amodei is not alone. Sam Altman, Elon Musk, and other leading industry figures have backed calls for greater oversight of frontier AI. The difference today is the unusual amount of agreement between companies otherwise locked in a fierce race to build more capable systems.

This phenomenon can be understood through an economic idea developed over a century ago by British economist Arthur Cecil Pigou regarding economic externalities. Pigou wrote as industrialization transformed Britain and the world, producing wealth alongside smoke and pollution—costs that did not fall directly on the companies creating them. According to Pigou, the private cost of doing something can differ greatly from its social cost.

AI faces the same problem, but on a much larger scale. AI leaders are warning that costs could be paid by humanity as a whole, ranging from job losses and cyber vulnerabilities to AI-assisted biological threats and conflicts. While the economic benefits in revenue, market share, and productivity are massive—particularly in science and medicine—the overall costs to society remain hard to calculate. Frontier companies are warning about risks that go far beyond standard issues like fraud, misinformation, deepfakes, and energy consumption, pointing instead toward biological and cyber threats, as well as increasingly autonomous systems.

Unlike Arthur Cecil Pigou's visible industrial pollution, AI lacks physical boundaries. Engineers cannot always explain why a particular capability emerges or predict every way a system might behave, making the question of liability and responsibility much harder to answer. AI companies appear to genuinely believe that some risks are too large for individual management, while recognizing that common rules protect them from solely bearing the cost of restraint. More than a century after Arthur Cecil Pigou examined industrial costs, we face the same dilemma: when private incentives and social risks diverge, who bears the cost, especially when the full social cost cannot be known until after deployment?