How Trump Was Persuaded to Regulate A.I.

I'm Tracy Bennett. I get to pick the word word every day, which is not as easy as it sounds.

“The fun fact about me is that I am descended from a witch who was put on trial in Salem.”

New York Times games are made by people, like the ones you just heard from. Go to nytimes.com/games to start playing today. Hey, it's Michael. Before we start the show, we want to let you know that our colleague, Pulitzer Prize winning Culture Critic, and host of the Times Podcast Cannonball, Wesley Morris, is going to be hosting a live event next week at the Tribeca Festival.

Wesley will be talking to the actress and activist Cynthia Nixon. They'll be having a conversation about great works of art about New York, which makes sense, given Nixon's starring role in arguably one of them, Sex and the City. It promises to be a very fun and very smart evening. The show is Friday, June 12th at 6 p.m.

And you can get your tickets at TribecaFilm.com/audio. Okay, here's today's show.

“From New York Times, I'm Michael Barrow. This is a daily.”

Today, the story of how President Trump, a grudgingly accepted that the most transformational technology of our time, artificial intelligence, requires government oversight. My colleague, Trip Nickel, takes us inside the dramatic White House battle that has played out over the past few weeks, and previews the coming fight over just how far federal

regulation of AI should go. It's Thursday, June 4th. Trip, welcome back to the daily. Thanks so much for having me. I want you to explain exactly what President Trump did a few days ago on Tuesday.

On Tuesday morning, as I was in the middle of my commute, I was bombarded with tax messages saying that President Trump had to many people surprised on an executive order to regulate artificial intelligence.

And it gives the government, for the first time, really, under the Trump administration,

oversight over AI models. This is a big change for an administration that has had an entirely hands-off approach to regulating the tech industry and the companies that are pushing artificial intelligence forward. What the government has done is said that AI companies will share their models voluntarily

with the government about 30 days before they release publicly so that the government can review the models and figure out if there are any software vulnerabilities that they could attack. And then that would allow the government to patch those vulnerabilities and prevent a cyber attack by a foreign adversary or just a bad actor.

This seems like a relatively big ask which is let us the government take a look at your software before the world, before all the businesses that wanted, get a chance to buy it. What's the story behind why this administration that has been so hands-off,

“decided at long last that it should be in the business of regulating artificial intelligence?”

Well, it's been on a journey. And that journey began last year when President Trump came into office.

And one of the first things he did was sign an executive order that repealed and

scrapped all of the rules from the Biden era that were designed to bring some safety processes to bear on artificial intelligence. And Trump would explain why he did this by saying, "Make America great again. We're going to make this industry absolutely the top because right now it's a beautiful

baby that's born. We have to grow that baby and let that baby thrive." The AI industry is a beautiful baby and you don't want rules to get in the way of a beautiful baby when it's taking its first steps.

You just want it to take its first steps essentially and let it start to run.

We can't stop it. We can't stop it with politics. We can't stop it with foolish rules and even stupid rules.

And that philosophy was really cultivated and encouraged by David Sachs, the White House AI's are. David is a venture capitalist, he's pretty much a libertarian, and he believes that any rules would get in the way of technologies advancement and development. And he wins the president's ear with two arguments.

If the US continues to lead an AI, we will remain the most powerful country,

but if we don't, we could fall behind our global competitors like China. On one hand, the geopolitical competition between superpowers between the United States and China, for who's going to control this technology that clearly has the potential to define the next century.

“And I think you could argue that the AI race is even more important than the space race,”

because it's going to determine who reshaped the global economy and who the superpowers of the 21st century are going to be. And on the other hand, he's also focused on the economic opportunity of this technology, and how can the United States harness that economic opportunity. There's a reason that our economy is driven by technology, and it's because of a largely hands-off approach during a previous era.

So given just how entrenched this thinking appears to be inside,

the Trump administration embraced by the president and the person that the president turns to for advice on the AI industry, what ultimately changes this thinking and leads to the executive order. In April, the entire tech industry got a wake-up call and so did Washington, and that is because anthropic, which is one of the leading developers of artificial intelligence in the country, announced that it had made a new model called Mythos,

that it said was very skilled at detecting and identifying vulnerabilities in software, and could be weaponized by people who wanted to do cyber attacks.

“And all of a sudden, everybody starts to think, "Oh goodness, like, what are the downside risk here?”

Like, what are the possibilities?" And what are the possibilities?

These are really powerful systems, and they can be used by bad actors to hack into the electric grid,

or hack into bank accounts, to steal money to all of a sudden overnight, turn off all the lights in New York City, and create mass panic. And Thropic said that it was so dangerous that it couldn't even release it to the public, and instead was creating a bespoke group of 40 companies that were going to get early access to this model, so that they could figure out how it might be used to find vulnerabilities in their software systems.

And concerns about this new model reverberate from the West Coast of the East Coast. First, you have Microsoft, which has written most of the software code that the US government relies on, calling the administration and saying, "Hey, this is a big deal. You need to take this seriously."

“And then on the East Coast, you have figures like Jamie Diamond putting calls into the administration”

and saying, "Look, this is a risk for banks. You all need to take this seriously." Jamie Diamond, of course, being the CEO of GPMorgan Chase, the Banking Giant. Right? And this gets the ear of Scott Bessant and Susie Wiles.

Scott Bessant being the Treasury Secretary, Susie Wiles, being the Chief of Staff. They worry that if there is a cyber attack, if there is some use of this technology in a way that shocks the American public, that if the administration continues to have a hands-off approach, it will look like it enabled that. And so they start to talk about, "Well, what could we do?"

Like, we're literally doing nothing now. How do we take a step forward of some sort and begin to get our arms around where this technology is going in the future? So all of a sudden, after this hands-off approach has persisted in prevailed, there's now a counter-vailing force inside the White House, Bessant and Wiles saying, "Wait a minute.

We could have a real problem here. Let's slow this down. Let's have a conversation about what actually a hands-on approach might look like. And where does that go?" In May, Scott Bessant and Susie Wiles begin to organize conversations within the administration about how it could regulate the industry.

And they begin talking to AI companies about what approaches might be okay with the company. Now, this is an administration that's trying to walk a tightrope here. I mean, they have said that we are very pro-innovation.

And so they don't want to fully abandon that idea.

But they want to also take a baby step forward or some step forward so that they do have some guardrails in place. And what they wind up coming up with is this idea of an up to 90-day window that they could review artificial intelligence models and assess whether or not those models are capable of finding software vulnerabilities.

In essence, it wouldn't tell tech companies what they can't do, but it would allow the government to absorb information about what their technology can do so that the government is prepared for that and can help other key groups like banks and utilities also be prepared for those risks. And so a draft of the executive order is written.

The administration schedules assigning ceremony in the Oval Office on Thursday, May 21. And they begin to ask and invite chief executives to come to Washington to be there for this event. And in the run-up to that event, a couple of things happened.

“Do you want to be now? Is White House National Economic Council Director Kevin Hassett?”

Kevin is great to have you this morning. Kevin Hassett, the head of the National Economic Council, goes on Fox TV. We're studying possibly an executive order to give a clear roadmap to everybody about how future AIs that also potentially create vulnerabilities should go through a process so that they're released at the wild after they've been proven safe, just like an FDA drug.

And describes the process that they're considering as something akin to what the FDA does for drug approvals. And FDA for AI, that would vet. You heard that correct folks, that would vet new models for safety. The thing we've been talking about not doing here.

This freaks out Silicon Valley constituents and leads to real worry that the administration is going to slow down the development of artificial intelligence.

And basically, just slam the brakes on it in the same way that it's like really hard to get a new drug out the door.

They fear that it's going to be really hard to get new artificial intelligence models out the door. And Thursday morning rolls around and it's time for President Trump to gear up to sign the executive order. And that morning, he begins to get skeptical about it and say, "What is this on my calendar?" And he calls, he starts calling. He starts calling tech executives and saying, "Hey, what do you think of this?" And so he speaks to a number of tech executives including Mark Zuckerberg, Mark Andrewson, the venture capitalist who has a tremendous amount of influence on the White House.

And most importantly, David Sachs. And David Sachs tells him, "Directly, don't sign this. This is a bad idea." And within a few hours of that, the President's on TV doing another of that. The White House today is an executive order on AI been postponed. Because I didn't like certain aspects of it. I postponed it.

“I think it gets in the way of, you know, we're leading China, we're leading everybody.”

And I don't want to do anything that's going to get in the way of... And he says that he's canceling the executive order signing a bit, because he's concerned that it's going to create rules that slow down the industry. I really thought that could have been a blocker, and I want to make sure that it's... This feels like a pretty decisive victory for the AI industry and for the David Sachs faction of those advising. The President, but we know that it's a temporary victory because Trump does end up signing some version of an executive order.

So after he changes his mind and cancels, the first one, what makes him want to sign it all over again?

There's persistence from Susie Wiles and Scott Besson that something still needs to happen. And over roughly the two weeks that follow, they continue to push this notion forward. And the President convenes a meeting this Monday at the White House with David Sachs. Scott Besson, Defense Secretary Pete Hecks up, and other principles who have a vested interest in this.

“And they basically wrestle with how do we get this done?”

And in the course of that conversation, there are a few changes made to the executive order. One is that 90-day window is reduced to up to 30 days. So cut down by a third, that's a lot. Yeah, that's a lot. And there's a line added about how nothing in the executive order can be construed to authorize a mandatory government licensing.

Those changes win David Sachs' blessing.

“And on Tuesday morning with zero fanfare and no advance notice to tech companies.”

The White House posted a notification on its website saying that President Trump has signed the executive order. And it is now policy.

Well, given how watered down the executive order ultimately becomes from 90 days to 30 days with all the stipulations of what it's not.

And given how quietly Trump signed it, how should we think about this regulation and how meaningful or not, it really is. It is small by design, no doubt about it, but it shows how the ground has shifted in Washington on AI policy. And the door has now been cracked open to have a bigger conversation about whether or not there should be other rules or regulations in the future. Remember, this technology is moving really, really fast and everyone was blindsided in April. So, you know, trying to catch up to where the technology is going is where a lot of people's head is right now in Washington.

We'll do a back. I'm Paul Centaurio.

I cover soccer for the athletic.

Whatever you call it, the biggest competition in the sport is happening right now.

“And the athletics World Cup coverage has everything you need to follow the tournament.”

This 48 country-staking part from the tiny island of Curacao to the five-time Champions Brazil. Even if you don't know your off-site from your on-site, if you eager to know more about the teams and matches, all the stories on and off the pitch, we've got you sorted. Maybe you're the kind of person who's already up early every weekend, making the neighbors win your favorite club scores.

We'll make sure you get equipped with more information more insight than anyone you know.

We've got more than 70 obsessive reporters on the ground covering the ins and outs from every game. I almost forgot to mention the best part, Amy. Free access to the athletics World Cup coverage in our app. Download the athletic app and see that. I want to turn to this bigger conversation that's happening in Washington about what the right level of regulation for AI is.

I mean, obviously the way that the fact that I was inside the White House for us all this debate was through this baby step of this executive order. My sense is that beyond the White House, there's less interest in just baby steps. Yes, I mean, and there has been a steady drum beat that has increased in volume over the past year. And it's coming from two sides of the country politically. It's coming from populist on the left, people like Bernie Sanders and populist on the right.

People like Steve Bannon, they would all like to see the government take a more heavy-handed role in regulating artificial intelligence. largely because this is a really disruptive technology. It's going to have huge ramifications for the country. And they believe that there's not being enough done to put the government at the forefront of where it's going. Well, let's start with the populist on the right and what's motivating their concerns about AI. Artificial intelligence in a free society are totally and completely incompatible.

Fall freaking stop. Steve Bannon has long been skeptical of big tech and a lot of the Braula Garks and Braula Garks as he calls them. It's not that I don't think AI has tremendous benefits, it obviously does. But the downside is so unlimited. Because he's really concerned that they're going to sell out the American public and rich themselves and they're going to do it without any transparency or any accountability.

“I think that artificial intelligence that we have to have even more than regulation.”

I think we need him as something like a Geneva treaty or some sort of arms treaty. This thing is very dangerous, very dangerous. And that's where he wants the government to step in and provide some rules for the road. And I am a more of a hardliner in this than anything else I've ever been a hardliner on. He hasn't been really explicit about what those rules should be, but he's definitely channeled what he believes is the anger building of working class people.

Right. And he ultimately worries that these AI brologarchs are going to force Trump to sell out his own base, because Bannon was at the heart of creating the Trump political message that appealed so powerfully to that forgotten American worker. Exactly. And the interesting thing about this is, it's not just an economic argument that's gotten momentum within the MAGA base. There's also a moral argument that's been building as well.

“I stumbled upon this by talking to a pastor in Texas, you know, in a conversation with this pastor, I was like, why did you become part of the AI opposition movement?”

And he started doing it because he met with a parishioner who told him a story about a friend whose marriage was on the rocks because he was having a relationship with an AI companion and it had completely destroyed and gutted the relationship he had with his wife. You know, and this pastor is concerned about how many other people might wind up in the same situation in the future.

Right. Basically, AI has a road to perdition. Kind of moral rot. Right. And like, what stress and strain does this put on kind of the moral fabric of America?

And this isn't an isolated thing. There were more than three dozen pastors who signed onto a letter with Steve Bannon urging the Trump administration to actually sign this executive order that we spent time talking about. They just want somebody to take the wheel and they don't want that somebody to be the tech industry that seems to have the presidents here. I wonder if these pastors and Steve Bannon are compelled by the executive order of the president signed a couple days ago.

“I mean, they feel like it's a step forward, but they'd like to see a bigger step taken. And when I asked one of the organizers of the letter, what would you like to really see happen?”

They said that one of the things that they're supportive of is a bipartisan piece of legislation that Senator Josh Hawley of Missouri advanced that would create a national AI system to evaluate more rigorously and like on a mandatory basis. And the official intelligence models before they released. Hmm. So something firmer than what the president is proposing, which is voluntary, a forced, basically pre-clearance review process in which the government determines whether something is safe enough when it comes to AI to be used by corporations and all of us.

Exactly. Okay. So what about the populists on the left, specifically Senator Bernie Sanders, who you mentioned a little bit earlier, what's he calling for?

First of all, it's worth noting that Bernie Sanders has just gone deep on artificial intelligence.

Claude, this is Senator Bernie Sanders. He's interacting and doing interviews with Claude, the anthropic system. I have been taking a hard look at how AI impacts American society in terms of economics and job loss in terms of the mental health of our kids in terms of what it will do to the political process. He's really trying to immerse himself in this technology and where it's going, and he's made two proposals.

“Given the speed at which AI is developing and the impact it's going to have on somebody aspects of our life, would you support a moratorium on the development of new AI data centers?”

One of which is a moratorium on artificial intelligence. And when I spoke to Bernie Sanders and asked him about this moratorium, which probably doesn't have a real shot of becoming law, he said one of the things that's kind of shocking to him is that given that AI and robotics are going to impact every man, woman and child in the country, you would think that Congress would be talking about this and yet the conversations about artificial intelligence aren't really happening on Capitol Hill. And so that's part of the reason that he's throwing his way around and trying to introduce legislation that really sparks conversation.

And the second proposal may be even more radical than the first.

But a president, the time has come for us to reclaim what was stolen from us. And that is that the United States should take a 50% ownership stake in all of the biggest artificial intelligence companies as they go public. I will be introducing the American AI sovereign wealth fund act. This legislation would give the public a direct ownership stake in the largest AI companies in our country.

And that the entire country should share the profits and wealth of those companies as they continue to grow and develop.

“Bottom line, the future of AI and the fate of humanity must not be decided behind closed doors by the wealthiest people in the world. It is our future. We must decide it. Let's get it done.”

That's about as far as you can possibly imagine from the president's baby step of give us 30 days to evaluate the safety of something before it goes out in the world voluntarily. This would be the government owning, controlling the industry night and day. It's a big radical leap from where the president is at the moment. And suffice it to say the AI industry not a fan of this proposal.

It was largely met with shock and Silicon Valley and ridicule from people who think it's basically unacceptable socialism.

“At this point, given how much of a desire we can see on both sides of the populist wings of the parties for regulation of AI, what is the industry actually willing to accept at the moment?”

The industry at the same time these conversations are happening seems to be recognizing that it has a big problem with the American public at the moment. That it's got a trust deficit. Pretty clear based on the polling data that's been coming out, based on the rise in AI opposition. And then based on a very violent act that happened here in San Francisco, a Molotov cocktail was thrown at Sam Altman's home, Sam Altman being the CEO of Open AI.

And that was a bit of a wake-up call for everybody in terms of making people recognize that developing this technology without any rules also has risk for the companies who are pushing it forward.

And so what you've seen in the wake of the executive order, and we're just talking like days, Michael, this is just a few days later, Open AI came out on Wednesday with its own post encouraging Congress to sit down and adopt more rigorous rules for the artificial intelligence industry.

“Now this is a company that has largely opposed rules up until this point, and that to me just speaks to the type of shift that's taking place right now.”

But is that a genuine request by a major AI company to be regulated? Is that just a public relations gesture? This is also an industry that spends a lot of money in American politics, almost all of which seems to be directed at electing candidates who will not want to be regulating the industry. It's a great question. As somebody who covers the industry, it's hard to figure out at times where Silicon Valley is going on this. And Open AI is a good example. It's pushing forward this notion that democratic governments, not private companies alone, should determine the rules and safeguards for the technology that it's developing.

At the same time, one of its founders, Greg Brockman, is spending money to try to ensure that the people being elected to public office support a relatively hands-off approach, if not a totally hands-off approach, to artificial intelligence. Right. I mean, I have the question of whether in the end what ultimately may lead to meaningful regulation of this industry is what leads to meaningful regulation of any industry, which is some sort of massive crisis. The two that come to mind are the FAA, the regulatory body over the commercial airline industry. That comes into existence after there are many plain crashes.

In the 1950s, the government says we need a regulatory body to deal with this. After the financial crisis in 2008, the US government wakes up and says we need a much more robust set of regulations around banks and derivatives and American finance. And so we get that regulatory scheme. Is that just at this point the most likely scenario in which we get meaningful regulation for AI? Something goes quite badly wrong. Given the paralysis on Capitol Hill, I think that's the right way of looking at this, that there there needs to be not just a mythos that isn't released publicly, but there needs to be a mythos that has been released and the lights need to go out in New York City or something like that.

We've been dealing with tremendous polarization as a consequence of that. There have been mental health issues for young people. There have been concerns about child sexual abuse material.

“Right, there have been crisis on top of crisis. If there's not been a single rule, right, that segment of the tech industry remains unregulated.”

And so I can't help but look at this and wonder, could this kind of nascent segment of the tech industry artificial intelligence also just run out the clock in perpetuity. Right, you're raising the real prospect that tech and AI made defy the history of crisis and regulation in this country that we've had with most industry. It has so far and it's got a really big bully stick that it's carrying around and that is how much it is driving the U.S. economy. And we're talking about a third to more than half of GDP growth at this moment is coming from artificial intelligence and the build out of artificial intelligence and that is a big reason that it took so long for this executive order to be enacted.

But at the same time, many of those economic benefits aren't being distributed broadly to everyone. I mean they're flowing to the few across a broad swath of the country.

What people are living with are anxieties about what the technology is going to do is it going to eliminate my job is it going to result in a data center and the backyard is it going to break up a family. Those pain points begin to bleed from the progressive populist wing and the mega populist wing of the country into a broader swath of the public. Then the pressure to take action and have real regulation will be impossible to ignore. A little trip. Thank you very much to appreciate it.

Thanks so much for having me.

We'll do a back.

“Here's what else you need to know today.”

On this boat, the age of 215 to the age of 28, the concurrent resolution is adopted. On Wednesday, the Republican controlled House of Representatives voted to direct President Trump to withdraw U.S. forces from the war with Iran or when approval from Congress to continue it. Four Republicans sided with Democrats in a sign of the growing bipartisan opposition to a conflict that is now in its fourth month. The vote is seen as largely symbolic because even if a legislation makes its way to Trump's desk, he's expected to veto it.

“Meanwhile, Iran fired a barrage of missiles and drones at two U.S. allies, Kuwait and Bahrain, undermining the claim made by the Trump administration that the war is over.”

The attacks in Kuwait proved deadly. Iran struck Kuwait's international airport, killing one person, injuring scores of others and setting an airport terminal on fire. Today's episode was produced by Caitlin O'Keev. Austin, Chothar Vady and Diana Winn. It was edited by Mark George and Annie Minoff and contains music by Mary Elizano, Dan Powell and Diane Waw.

Our theme music is by Wonderley. This episode was engineered by Alyssa Moxley. That's it for Daily. I'm Michael Barr. See you tomorrow.

This week on the Booker View podcast, our monthly Book Club meets to talk about Ben Lerder's new novel Transcription.

It's really really a bad part.

“His 2014 book made the Times's best hundred books in the 21st century list.”

So whenever Ben Lerder puts out something new, it's an event and it's something that needs to be discussed.

We could talk about this book all day.

Listen to the book review wherever you get your podcasts.