The prevailing wisdom about American regulation of AI echoes Mahatma Gandhi’s famous quip about Western civilization: It would be nice if it existed. Whether or not Gandhi was correct, the dominant view of AI regulation in the United States is off the mark. It is simply not true that the U.S. government is failing to actively shape the way in which AI is used. Rather, American regulation is unfolding in ways, and in venues, that commentators have largely ignored. As the pace of such regulation increases, and starts to have spillover effects, the value of taking a wider-lens view is increasing.

The dominant view of American AI regulation focuses on Congress. It bemoans the American legislature’s failure to enact comprehensive rules for the field as a whole. On this account, there may be a multiplicity of bills percolating in the House and Senate—covering consumer protection, social media, government procurement, and more—but these have not gone anywhere, and will likely go nowhere, as a result of partisan deadlock. The executive branch under both Presidents Trump and Biden has, to be sure, generated executive orders and (in the second case) a “Blueprint for an AI Bill of Rights.” While these have catalyzed useful additional policy documents from the National Institute for Standards and Technology (a risk management framework) and the Office of Management and Budget (on government AI), these hardly cover the field of AI use cases, but are rather second-best outcomes. Precisely because AI is so flexible and so extensive in its potential uses, such narrowly-tailored interventions seem far from adequate to the regulatory task at hand. In contrast, the European AI Act looks set to monopolize the regulatory field globally thanks in large part to its first-mover advantage. 

Read more at KU Leuven AI Summer School Blog