Last Tuesday, I wrote about the ideology driving the AI doomers. I tried my best to take their fears seriously, since I share some of them myself. Even if you haven’t read that one yet, you’ll probably be unsurprised to learn that I landed where I usually land: a decentralized, competitive market is the only safe path. Centralized “solutions” lead to tyranny every single time.
To those of you worried about AI, that might sound more like blind optimism than a real solution. It’s one thing to say the market will handle it, and another to actually show how.
Coincidentally, the very day I published that essay, a more fleshed-out answer was being drafted at the White House. The Accord on Super Intelligence, signed by President Trump, the CEOs of Anthropic, Google, Meta, and Nvidia, the president of OpenAI, and Elon Musk, laid out a framework for “self-policing” within the industry.
The Not-So Voluntary Accord
On Friday’s episode of the All-In podcast, former White House AI czar David Sacks broke down what exactly went down during last week’s AI summit. As he reported, Mark Zuckerberg and Jensen Huang proposed the four-part framework that the group ended up voluntarily signing onto.
First, each company will establish “robust internal controls” regarding AI safety. Second, an internal team will oversee the implementation of those controls. Third, an “independent external auditor” will validate their internal control systems. Finally, an “independent committee of the board of directors” will receive and act on reports from the auditors.
Still, this might sound to you like an overly complicated way of saying, “Don’t worry. We promise we won’t do anything bad.” But, as Sacks explained on the show, signing onto an agreement like this comes with real consequences. The FTC and the SEC have “the ability and the responsibility to enforce the public commitments of these companies.” If they don’t follow through on the terms of the agreement, each board risks losing its Directors and Officers liability insurance, opening them up to all sorts of problems when something goes wrong.
President Trump described the Accord as “morally binding” when asked about enforcement, but frankly he was underselling how effective this approach could be. I’m sure there are lots of lawyers ready to pounce if any of these companies drop the ball. The stakes are high for everyone involved.
We’ve Done This Before
If an insurance policy doing the work of a regulator sounds like a clever new idea, it isn’t. It’s how America has kept dangerous new technology in check for more than a century.
The progressive story about safety goes something like this. Left alone, companies race to the bottom and cut every corner for profit until their misdeeds are exposed and the government steps in to save us.
That story ignores the actual history. In 1893, MIT-educated electrical engineer William Henry Merrill was tasked with inspecting the electrical installations at the Chicago World’s Fair for fire risk on behalf of the Chicago Fire Underwriters’ Association. The following year, he opened a lab to test the safety of electrical products and fire protection equipment.
In 1906, the lab started labeling products that passed its tests. You’ve almost certainly seen it. Check your refrigerator or your microwave and you’ll probably find two letters in a circle: UL, for Underwriters Laboratories. The underwriters in question are insurance underwriters. If you’re going to insure against risk, you need to actually know what the risks are. UL provides that information.
Nobody passed a law to make any of this happen. The people who would pay for the fires had every reason to prevent them. Manufacturers wanted the label because customers and insurers trusted it, and insurers trusted it because UL’s entire business depended on never rubber-stamping junk.
It’s not a theory. It’s not wishful thinking. It’s a chain of incentives to not kill your customers or burn down their homes that’s doing the work.
Problem Solved?
UL isn’t just a historical model for free-market AI regulation. It’s already on the case. In 2025, UL Solutions, the for-profit arm of the original lab, launched its first AI safety certification.
Progressives insist that the only way to hold a company accountable is with government force. They dismiss market discipline as “self-regulation.” Of course, President Trump didn’t do himself or AI companies any favors by referring to the Accord as “self-policing,” but that doesn’t change reality.
Mr. President: Next time you host a summit on AI safety, invite UL too. Maybe the left will learn a thing or two about markets.