Here’s a brief history lesson. In 1966, nearly 51,000 people lost their lives on U.S. highways. Evidence showed that many of those deaths could have been prevented with modern seat belts, which had already existed for several years. The automotive industry responded by voluntarily agreeing to install the safety feature in every new vehicle—a self‑regulated triumph.
In reality, that never happened. Instead, Congress created the Department of Transportation in the fall of 1966 and enacted the National Traffic and Motor Vehicle Safety Act together with the Highway Safety Act, granting the federal government broad authority to set safety standards. By 1968, seat belts became mandatory in all new cars. Despite a dramatic rise in highway traffic over the past six decades, road fatalities have declined significantly.
I recalled this history while observing a group of AI CEOs gathered with President Donald Trump on Tuesday, congratulating one another on signing an AI safety agreement. My colleague Maddy Varner detailed the accord’s provisions elsewhere; the essence is that compliance largely rests on the AI labs themselves.
As Trump put it succinctly in the Oval Office on Wednesday: “They’ll police, they’ll police each other, they’ll self‑police, they’ll police themselves. It’s going to work out very well.”
That captures today’s state of AI safety regulation. On one side stand AI executives, employees, and tech leaders who call for clear, thoughtful, and actionable rules to avert potential catastrophe—Bill Gates, for instance, has been warning of an AI apocalypse in recent media appearances. On the other side lies Trump’s belief that repeating a statement enough makes it true.
None of this is simple or easy. The challenge of how to properly “pace the frontier,” a phrase favored by those same executives, remains unresolved. Even if the major AI labs agreed to slow development, enforcing such a pledge would be difficult, and China would likely pursue its own trajectory regardless. The U.S. economy increasingly relies on the AI sector, which has evolved into a self‑reinforcing loop of financing and circular investments; any serious crackdown could trigger instability far beyond Silicon Valley.
The automotive analogy is imperfect for many reasons. There is no seat‑belt equivalent for AI—no simple fix that can demonstrably cut risk overnight. (It also took decades for most states to pass laws requiring seat‑belt use.) The potential harms of AI are less certain and harder to quantify. Doomsayers warn that the technology could cause human extinction within a decade, while skeptics dismiss such claims as science‑fiction hype.