Remarks: JD Vance Holds a Firechat Chat with Joe Gebbia on AI - September 29, 2026

Thank you.

All right. Well, VP, welcome to Hello America.

Yeah, it's great to be here.

[Laughs] It's incredible.

Congratulations.

Thank you.

I know you guys were working very hard on this thing down the stretch, so congratulations. It's, it's, it's an amazing thing to see it get off the ground.

Thank you so much. Uh, it was a great effort this morning and team's been, uh, working away. But, uh, great to have you here. And, um, I thought we would draft off of the conversation that was just taking place and, um, you know, we were just coming from a, an event over at the White House.

Yep.

Talking about that topic with the leaders of the industry. Um, so I though it'd be interesting to, to share a recap with the audience. Ob- -- observations, perspectives that --

Sure.

-- came out of the lunch that you can share, of course.

Yeah, yeah, yeah. Well, first of all, Joe, congratulations to you and your team again. You know, when, when Joe and I first started talking about this, the one thing I said is, "Don't let it be a clusterfuck like America.gov." [Laughs] Hopefully -- oh, oh, sorry. Excuse me. Hopefully no kids in the back.

Uh, or excuse me, like healthcare.gov. And America.gov has gotta be the good version of that. And I think you guys have done an amazing job. So give it up to Joe and his team. An amazing thing. You know, it, it is, it is possible actually to build good software tools for the American people. You just have to have competent people doing it. That's exactly what you've done.

So lunch today, obviously, you know, be a little bit, uh, a little bit, um, careful about it because it was a private conversation. But, you know, what the president wanted to do is obviously be responsive, first of all, to the American people's concerns because you have a lot of doomerism about AI or super intelligence as we're now calling it. And there's a lot of, you know, what's, what's actually happening here.

And I think a lot of Americans, when they think of AI or SI, they think, "Oh my God, this is Skynet," right? They think the terminators are gonna come and take over everything and, you know, take over our nuclear weapons and have the robots running around killing people. And that's just not gonna happen.

That's actually not gonna happen. Now, what could happen is you have people who are building AI tools, not actually being careful about what they're doing, about the responsibility they have to build products that actually work, that don't hurt people, et cetera, et cetera. And I think what the president wanted to drive home was, y- -- you know, we need to actually have a AI conversation that's built around making sure these tools actually work, but not absolving companies of their responsibility to build things that are actually safe and actually work.

And there's been this weird dynamic where a lot of people who are developing the frontier models in artificial intelligence, they'll come out and they'll say, "Oh, oh my God, this is a terrible, terrible thing we're building." It's like, "Well, if you're building a terrible, terrible thing, then stop and build something good." Don't come to the government and regulate.

It, it's, you know, it's, it's like a kid who, who just can't stop eating cookies. And it's like, I just, "Please, government, come and help me stop eating these cookies." No, no, no. You have to exercise some responsibility over your own products.

[Inaudible]

And I think that that is the first --

The cookie jar up to the top of the shelf high?

Exactly.

Okay.

Exactly. And, and, and, and the, the first step here is just to actually, you know, you hear this term alignment. I think the most important alignment is for the frontier models to realize that they have to build products that are actually safe and good for American consumers. And if they understand that the government actually has pre-existing laws on the books where if you build something that gets unleashed on the internet that is used as a tool for cyber warfare, then you have responsibility for the products you develop in the same way that a pharmaceutical company has products liability for unsafe pharmaceutical products.

Or in the same way that, you know, if you build a, something that, that plumbers use to fix people's houses and it doesn't work, you actually have responsibility. And I think reemphasizing that important primacy of product liability, making sure your products actually work and taking responsibility over these tools is very, very important.

Now, obviously there are risks to AI, and we talked about that during the lunch today. But I think the president wanted to say, "Look, the solution to some of the AI risks is for you guys to take the risks seriously, not to come to the government for a regulatory regime that may make things worse if it's not smart and careful," which is exactly what the president wants.

Mm-hmm. And in that, that metaphor of product liability, um, what, what is the group that oversees when somebody makes a violation, FDA or, or similar in your mind?

Yeah. I mean, I, I think the obvious, you know, the two obvious agencies are the FTC, the Federal Trade Commission, and also the Department of Justice, right? Uh, if you create a product in the United States today that violates some right of the consumer, that actually doesn't work as advertised, that causes some harm, then we have, again, pre-existing regulatory agencies.

I, I've heard some talk about, you know, do you need an FDA or an FAA for new frontier models? And I think that, that the problem with that is, one, you have to get regulators who are actually smart about what's going on. And most bureaucrats just know way less about this, uh, than- than, uh, the people who are actually building these products.

So it -- it -- it's -- you can't just create that solution out of thin air --

Mm-hmm.

-- and hope that it's actually going to work.

Mm-hmm.

You have to be smart and deliberate about it. And one of the things I know the President's very worried about, I'm certainly very worried about, is you see talk about risks, and I think that's good. You, obviously, want to have alignment in these models, but that conversation about risks almost immediately translates to, here are a whole host of regulations that we should throw at the problem.

And I'm a strong believer that regulations can actually make us safer and they can be consistent with innovation, but only if they're careful and only if they're deliberate. And I don't think anybody regulates well in the middle of a panic. COVID should have taught us that.

Mm-hmm. Mm-hmm. Uh, and- and if you think about the- that group that, if it's an FDA or- or other, uh, if they're not from industry, they may not have the insights to actually guide things in the right way. How do you find that objective group of people? If they're from industry, there's maybe some bias.

If they're not from industry, they don't understand technology.

Yeah. I, I guess I -- what I think is, I'm a big believer in competition and I'm a big believer in strong antitrust laws. Because you're exactly right, some people from industry are going to be biased. Obviously, I think the nature of running a company is you're going to be biased towards your company. And if you're not, you're probably in the wrong business.

So how do you balance obviously the fact that the people who know the most about the industry are also the people with the strongest financial incentive to not maybe care about safety as much as profit or some other thing? And I think the answer is competition, because you have a number of companies who actually know a lot about this industry, who know a lot about the technology, but they don't have the, maybe the same fin- -- financial incentives as the Frontier Labs.

Or, you know, let's take one obvious example. I mean, you know, there's a lot of concern. I think the most obvious negative use case of the Frontier models up to this point --

Mm-hmm.

-- is the cyber capability. We have seen some of these models, by the way, they've been trained this way, but they have remarkable ability to enable hackers to hack into systems. Okay, so the obvious answer to me is to say to all of the technology companies that might be subject to a model run amuck in a cyber sense is to give them the tools to actually patch their code and protect themselves against cyber attack.

In other words, the best weapon against cyber offense is good cyber defense. And the companies that don't want to have their companies or their, their websites hacked into, those are the companies who are best situated to do this. So I think so long as we encourage healthy and robust competition and so long as we remind companies that they actually have responsibility to build products that work and products that are safe, I think that is, actually, that goes a long way to fighting the battle.

What I think would be a huge mistake is to give, for example, Section 230 protections --

Mm-hmm.

-- to some of the new technology companies. That actually shields them from liability rather than making them more responsible. And I think it would be a huge mitake to allow- -- mistake to allow cartel-like activity. We don't want the four or five biggest model companies, and some of these guys, these companies, by the way, are run by very smart, earnest people, very good people, I have nothing against these companies, but I want those companies to compete against one another, not to join in a sort of unified cartel against the American consumer.

They protect themselves, but I think it would be bad for the American people.

Yeah. How do you make sp- -- space for the little guy in all this?

Exactly. Exactly. And I, and I think, you know, like I, I have- I have different friends who are on different sides of the open source debate.

Mm-hmm.

And, but to me, you have to protect competition.

[Inaudible]

You have to make it possible for a young upstart to, to actually enter these industries, to compete in these industries. And this is where I think robust antitrust protections and also robust product liability protections --

Mm-hmm.

-- these things already exist. And, and maybe, like, to be clear, maybe we find out that a year down the road or two years down the road, these existing regimes are, like, inadequate to the challenges.

Mm-hmm.

But let's actually try what has worked in the past. And if it doesn't work in the future, then build it out from there. Let's not just throw a massive new regulatory regime against a problem that could make it worse.

Amazing. Well, thank you so much for sharing your thoughts on AI.

Yes.

Um, I want to switch gears real quick because, uh, there's another part of your life which has to do with fraud and, um, so I'd love an update because I, I was so excited when I heard about the task force that was set up. I got to meet with some of your team members, um, early on. And it was like, by God, there's, there's really, you know, a bigger team thinking about a massive issue in our country.

And I, I'd love for you to share with us some of the, the insights --

Yeah.

-- findings, observations, surprises --

Yeah.

-- through that task force.

Well, obviously, you and Elon were doing amazing things at DOGE, and what we wanted to make sure is that this effort to root fraud out of our government didn't just get abandoned because, you know, the DOGE, the DOGE effort fundamentally had a short-term sort of lifespan because as much as we might have loved it, Elon was not gonna stay in the federal government for the rest of his life.

His shareholders would not have loved that.

[Laughs].

So what we wanted to do is to make sure the anti-fraud effort still had teeth and was still very much a focus of the administration. The, the one thing I've learned about fighting fraud is how much it has to be an interagency effort. And so, you know, we, we find, for example, that, you know, you have an illegal immigrant who's committi- -- committing SNAP fraud.

So they're taking advantage of the food stamps program. You actually can't create proper enforcement mechanisms unless you have the Department of Justice coordinating with the Department of Homeland Security, cobor- -- coordinating with the Department of Agriculture because the Department of Agriculture administers these programs.

So what, what we've tried to do is just take a whole of government approach to it. And obviously, like, I'm empowered to do that because I'm the vice president of the United States. And it's, it's been amazing to see everybody work together. I know you spoke with Oz earlier today. Dr. Oz is an amazing guy, incredible energy.

But some of the best fraud fighters have been people who work for Dr. Oz who saw, "Oh, wait a second. There's this part of the Medicaid program or part of the Medicare program that isn't actually giving healthcare to Americans who need it, it's actually giving a lot of money to fraudsters who aren't even providing the services they claim that they're providing.

Let's just cut that money off and actually save the American taxpayer a lot of money in the process." So it's been, it's been very rewarding work. I've had a lot of fun being vice president --

[Laughs]

-- but that's, that's at the very top. And, you know, Joe, to, to connect this back a little bit to some of the stuff that you're working on and your team is working on, the biggest challenge that we have in fighting fraud is that nobody has a, like, good, consistent, coherent, and, you know, broadly applicable way of verifying eligibility for these things.

Mm-hmm.

And so the SNAP program, you apply for it in California, they won't even, like, check your identification. Like if I go to the state of California, we've in fact had this conversation, and I say, "Who is it that's on your welfare roles in California?" They can't even tell me the answer because they don't do the basic work of identifying whether the people receiving these benefits are eligible to receive those benefits.

Now, that sounds insane and crazy, but like hundreds of billions of dollars, maybe trillions of your tax dollars, every single day go to places where we don't verify that the people receiving those benefits are actually eligible to receive them. Where I see like this America.gov thing coming full circle, and again, like I played around with the website a little bit, it's amazing.

It is the exact opposite of healthcare.gov, which I'm very proud of. [Laughter] Um, but I'm sorry I said the F word to make that point, but --

[Laughs]

-- hopefully the TV cameras won't penalize me. But here's the thing, if we could just have a smarter way of verifying identity and eligibility for these programs, we could save the American people a lot of money and importantly, we could ensure that the people who are receiving these benefits actually deserve them.

Because one of my big beliefs, Joe, on this is like the, the social welfare system in this country is a core part of the social contract.

Mm-hmm.

Like my family, I myself benefited from some of these programs. I am now the Vice President of the United States. Uh, I know a lot of people who get down on their luck, who need a little bit of help, you know, giving their kids access to a doctor or taking their kids to a dentist or need a little bit of help that month, you know, buying food for their families.

I want those programs to exist and I think it's a cool thing about Americans that we're generous to one another. We say, when you're down on your luck, yes, of course we want you to be finding a job and yes, we want you to be trying to get back on your feet and we know most of our fellow Americans will, but if you're down on your luck, your kid is not gonna go hungry or not gonna be able to treat an illness because they can't afford a doctor.

But that social trust, that generosity rests on a foundation where you write your check to the IRS and you know the IRS is gonna be careful about where it goes. That has not been true for decades in the United States of America. The President of the United States said, "Change it. Make sure we know where our money's going." And that's exactly what we're trying to do. And I think you guys are gonna help us.

I can't wait to do it. Ladies and gentlemen, let's hear it for Vice President.

Awesome man. Thank you.

Yeah.

Thank you, man. That's great. Beautiful.