How Do You Regulate an Industry You Don’t Understand? | Columbia Magazine

Columbia Magazine has asked the question that has been sitting in the corner of every AI policy panel for three years, and asked it plainly: how do you regulate an industry you don't understand? The answer, per the excerpt, is that you mostly don't. President Donald J. Trump scuttled the plans Joseph Biden set in motion in 2023 — federal safety standards for AI, plus government-overseen testing of models before release. Two years of drafting, a whole apparatus of clipboards and pre-deployment forms, and it went out with the same bureaucratic shrug that greets everything in Washington that hasn't got a lobbying budget.

Into the vacuum steps the New York Times editorial board, which the piece notes has advocated a federal agency to oversee AI. Splendid. A federal agency, presumably staffed by the same people who brought you the DMV's website, tasked with understanding a technology whose own creators describe its behaviour as emergent. The industry, meanwhile, has already supplied the answer it prefers: companies need not wait for regulations to decide what safety guardrails to adopt. Read that twice. It's not a commitment, it's a shrug in a blazer — the same self-assessment arrangement we looked at last week, in which the lab grades its own homework and files the note with nobody in particular.

So the grand settlement is this: the pre-release testing regime is gone, the agency is a talking point, and the safety standards are whatever the company feels like on the morning of the launch. The title asks how you regulate an industry you don't understand, which does rather assume someone intends to try. The likelier outcome is that we keep asking the question in glossy magazine spreads while the guardrails remain a matter of executive mood and quarterly optics. You don't regulate what you don't understand. You just hope the people who built it are having a good quarter.