Make AI Labs Pay For Mass Damages | Judge Glock

Judge Glock (0:00)

If these products at some point become so dangerous that they could do mass, civilizational style harm, then yeah, then we're just kind of in the general territory of bands.

But I think that's an unlikely scenario. The railroads kill thousands and thousands of people each year. Cars still kill tens of thousands of people each year. We don't shut cars down, even though they kill tens of thousands of people a year. What we should do is make sure the companies are gradually internalizing the harms they're imposing on society.

SPEAKER_2 (0:35)

All right, we are back. We're live with Judge Glock, who is the Director of Research and a Senior Fellow at the Manhattan Institute. And he's recently written a very good piece about tort law and AI. So Judge Glock, welcome to MTS.

Judge Glock (0:50)

Thanks so much for having me.

SPEAKER_2 (0:51)

Absolutely. So tell us about your proposal for how tort liability around AI should work.

Judge Glock (0:56)

Well, a lot of the proposal is actually just to rely on the existing systems of tort in America. So tort law, for those of you not familiar with it, is basically the law that allows people to sue companies or other individuals for harming them. If you get hit by a company's truck, you can bring a tort law claim against the trucker or the company to have them make you whole as it's known in the law. Basically, make sure that they pay for your medical bills, any lost work or any damages they cause to you and your life.

So tort law is one of the main ways we've regulated businesses throughout American history. We don't think about it as much because we obviously now have a massive regulatory state that designs to do this through bureaucracies and through former rules and rulemaking and federal litigation by the government. But one of the main ways we've regulated business and especially new technology is through this tort law. If you, say, built a new railroad in the 19th century and you hit a bunch of cows along the way, or you happen to have some embers from your railroad go out and burn a cornfield, you could be potentially liable, you as a railroad company there. And what I was saying here is that we should think about the harms from AI companies in a similar way. I'm not saying that's the only way to deal with harms from AI companies. But ideally, we want them to not think about trying to satisfy some federal regulator or some pseudo expert body that gets set up to decide what kind of model or proposal would or would not make sense, but make them pay for actual harms they do. If they bring down a power plant and knock out a lot of a city or something, they should have to pay for that. And tort law is a way ideally to internalize the harms that and potentially be caused from AI without having a top-down, bureaucratic, regulatory body administering it.

SPEAKER_2 (2:59)

Is that actually possible? It seems like in almost every field of law, a top-down regulatory body has emerged alongside existing tort law. Like even with social media, we had Section 230, which provided a pretty broad liability shield for a very long time. And now it seems like that is about to be overwritten by new laws and court rulings and regulations.

Judge Glock (3:23)

And obviously, it also hasn't stopped a lot of damages claims against these social media businesses.

I forget how much exactly, but was it $17 billion that Metta slash Facebook had to pay recently because of supposed harms to their users?

There's a lot of tort law going on in America today. Yes, as you say, often in addition to these federal regulatory bodies, but often also as the main source of regulation. So I would draw to your attention the Consumer Product Safety Commission. This is a 1972 commission set up during the Nixon administration that's supposed to monitor consumer product safety. But almost nobody has heard about it. And the main ways we regulate consumer product safety today is still through the tort system. You put out a toy that blows up and hurts some kid and their family, you the company often have to pay for it. If anything, I think the legitimate concern about the tort system is that it's been too invasive and too extreme in terms of the penalties that's imposed on companies. But that same feature should comfort at least to some extent some of those AI do-mers or those concerned about AI causing significant amounts of harm. The tort system can be very, very extensive in the types of damages it demands from companies.