
Loading summary
Unknown Speaker 1
Okay, don't laugh.
Alex Boris
I'm gonna laugh.
Unknown Speaker 1
I bought a domain at 2am last night.
Alex Boris
I love that for you.
Unknown Speaker 1
It was just there calling to me.
Ezra Klein
Where?
Unknown Speaker 1
On wix. I even built a whole website last night with its AI website builder WIX Harmony. But I was still in bed by 2:30, so you should be proud of me.
Ezra Klein
That was very responsible of you.
Unknown Speaker 1
You should try it too. Maybe I will just go to wix.com
Ad Voice 1
domains whether it's a birthday trip, a
Ad Voice 2
family reunion or just a fun getaway, booking a VRBO vacation rental means no worrying about surprises. VRBoCare and 247 Life Support have your back if something's off. The Loved by guest filter helps you find top rated homes. And verified reviews mean real feedback from real VRBO guests so you know exactly what you're booking.
Ad Voice 1
Honestly, I just booked my VRBO because there was a sweet wine fridge.
Ad Voice 2
Hey, we all have our reasons. Don't walk into a surprise if you know you verbo terms apply. See vrbo.com trust for details.
Alex Boris
The super PAC that's attacking me that has $140 million and was spun up by Andreessen Horowitz, which is now the largest spender in campaigns in the country, period, is focused on ensuring that there is no regulation of AI whatsoever. And frankly, they don't have to win forever. They just need to win for a couple election cycles because the speed with which AI is increasing its power and therefore they're increasing their wealth and they're increasing their power, is unprecedented. And so they've decided to spend $10 million against me in this race, which is kind of a compliment. But they've made that investment and they've said publicly to make an example out of me, to intimidate anyone else in Congress from wanting to regulate this at all. And, and if they win, it's gonna send a really loud message that if you wanna keep your job as an elected official, don't actually pass a bill in this space.
Jon Favreau
I'm Jon Favreau and you just heard from today's guest, New York Assemblymember Alex Boris.
Ezra Klein
Alex is running for Congress In New York's 12th, the very competitive Democratic primary that includes candidates like the Bulwarks, George
Jon Favreau
Conway and Kennedy Air Jack Schlossberg. But aside from the high profile names,
Ezra Klein
there's another story playing out in the race and it's about AI.
Jon Favreau
Specifically, it's about pro AI super PACs
Ezra Klein
funded mostly by OpenAI and Andreessen Horowitz, spending millions of dollars trying to defeat Alex. When you Talk to Alex.
Jon Favreau
Those attacks might seem counterintuitive. He doesn't consider himself anti AI. He doesn't want to shut it all down.
Ezra Klein
But he's also been an outspoken advocate for the regulation of AI.
Jon Favreau
He was the author of the Raise Act, New York's landmark law that sets transparency requirements and safety standards on AI developers.
Ezra Klein
And he's released one of the most sweeping AI policy frameworks for how Congress should regulate the industry, which is why some of the AI companies are spending so much money to defeat him. I invited Alex on to talk about what AI regulation could actually look like
Jon Favreau
if we had a Congress and a
Ezra Klein
president willing to get it done.
Jon Favreau
We talked about how to guarantee the
Ezra Klein
dignity of work in an age of full automation, how to protect our kids
Jon Favreau
from the dangers of AI, and why he thinks that the wealth AI creates
Ezra Klein
should be redistributed to people as an AI dividend.
Jon Favreau
I also talked to him about the
Ezra Klein
politics of AI, both the populist anger
Jon Favreau
against the technology and the let it
Ezra Klein
rip approach of the AI companies and the Trump administration.
Jon Favreau
And of course, we also talked about
Ezra Klein
Palantir, his one time employer that's currently
Jon Favreau
running ads in the District attacking him
Ezra Klein
for working at Palantir. We'll get to that conversation in a moment.
Jon Favreau
But before we do, if you're a friend of the POD subscriber, which if you aren't, you should be, you can now buy tickets for this year's Crooked Con special presale just for subscribers. But if you're not a subscriber because you hate pro democracy independent media and
Ezra Klein
you love listening to podcast ads, you
Jon Favreau
can still buy Crooked Con tickets next
Ezra Klein
week starting on May 19th.
Jon Favreau
Either way, it's gonna be a big fun party right after the midterms November 5th through 7th in Washington, D.C. so go to crookedcon.com for more details, including
Ezra Klein
how to become a friend of the POD subscriber. All right, let's get to it. Here's Alex Boris. Alex, Boris, welcome to Offline.
Alex Boris
Thanks for having me.
Jon Favreau
You have a fascinating background.
Ezra Klein
One of your earliest memories was being on a picket line with your dad when you were 8 and he was locked out by Disney over health benefits. You studied labor relations at Cornell, did anti sweatshop organizing against Nike, and then at 22, you went to work for Palantir. Bit of a zag. I know you've said you enjoyed economic modeling and working with data, but why Palantir?
Alex Boris
It's a great question. I always thought I was going to be a lawyer, but every lawyer I spoke to Told me not to be a lawyer or at least take time off in between. Yep, yep, common advice. And at the very least, take time off in between and make sure that's what you want to do. And so I was at a economic consulting firm called Cornerstone Research that prepares expert witnesses for trial so I could interact with lawyers and see them. And as you referenced, I found I really enjoyed economic modeling. I really enjoyed playing with data. I did not enjoy the billable hour and the doc review. And so I was like, okay, everything I like about this job I don't get to do as a lawyer. And everything I hate about this job is exactly what I will do as a lawyer. So let's lean in on the data side. And then the question was, you know, what kind of data? Where do you want to go? And I actually, I had an offer from a startup that was just using data to help kind of predict future trends and sell into finance. But for me, government was always part of the appeal. I am a Democrat and so I believe that government can and should be a force for good. But I think that also gives us the burden of proving it. And I went to Palantir in 2014 because they worked with government, because it was a chance to prove that government could actually work and spend my time on federal civilian projects where I got to live that out.
Ezra Klein
So thinking back on the pitch from Palantir back then that the company existed to use technology to prevent fascism, what do you make of that pitch now? And do you think it was always bullshit? Or do you think that Alex Karp and the company changed?
Alex Boris
I don't know that everyone at the company believed it the full way, but I don't think it was fully BS either. I think that was certainly among many of the employees. I mean, many of us that were hired in really thought that was the mission. And I don't know Alex Karp particularly well. I haven't said a word to him in seven years. But at the time, and my 23 year old self believed the pitch and believed that that was the focus. It was also, you know, this was during the Obama administration. This was very much an optimistic time of we can solve problems if we just put the right people together, the right software together, we design it the right way. And frankly, a lot of the work was in line with that. I mean, I worked with the Department of Justice to hold the largest banks accountable for their role in the Great Recession. While I was there, we recovered $20 billion for taxpayers. I helped the VA better staff their hospitals and provide Veterans care that they need. I helped the CDC in a very dorky project, but. But I like to nerd out. We helped the Census Bureau and Bureau of Economic Analysis slightly improve the accuracy of gdp. It's a rounding error over a rounding error. But there's a great paper I'm proud of with my name misspelled in it. And so it was work to just make government work. Now, we were told there were also guardrails. There were ethical guardrails. There was guardrails we were building into the software itself and those were real audit trail access controls. And there were also guardrails on who we would work with. We weren't going to sell to the Chinese government, to governments that are antagonistic to the US and our allies. But then there's this third level, which is contractual guardrails. Hey, you cannot use the software for A, B and C. And I don't think it was a BS versus not bs. I think it was inconsistent. At the doj, we had those guardrails. When Trump came in, leadership at the DOJ said, hey, we want to use the software for immigration. And I, as the lead of the project, said no. And I was fully empowered to do that because the contract was written in such a way that it was three mutually agreed upon case types. But on other projects, that was not what Palantir leadership chose. And in fact, that Palantir allowed the Trump administration to use the software for deportations is why I ended up leaving.
Ezra Klein
So you leave in 2019, three years later, you're in the New York State Assembly. Why elected office and not a nonprofit, A startup, something like that, really an experiment.
Alex Boris
I went from Palantir to a startup that was doing anti money laundering counterterrorist financing, working with really early AI models for those that want to geek out. Bert and Laser for the two listeners that will recognize that. And. And then I went to one called Promise, which helped to distribute aid that municipalities and states had allocated, but wasn't actually getting to the people that needed it. And then my assembly seat opened up. And it wasn't an obvious, oh, run for it. But I had conversations with friends and family, one of whom was saying, listen, you're always talking about how you're downstream of bad policy, trying to fix it with tech. Here's your chance to go upstream and design it right in the first place. And, and second of all, run for it. You don't know if you're going to win and if you do win, but you hate it. It's all politics. It's all mudslinging, you can't actually get anything done. Then in two years, in four years you quit and you go back to what you're doing now, but you can't in two to four years, say now I'm going to run for the open seat. Like this is a moment in time and an experiment. And I ran very much with that attitude of, hey, can we prove that government can actually deliver for people?
Ezra Klein
So before we get into what you want to do about AI, I'd just love to hear your big picture view of where the technology is, where it's going. What are you most confident about for the next five years with AI?
Alex Boris
That five years is way too long a timeframe to predict.
Ezra Klein
Wow. The thing that I maybe we should do one year, two years.
Alex Boris
Yeah, yeah. No, I think the thing I am most confident about is that this technology is moving faster than any in human history and will likely only accelerate it. That not only because it's the technology that has the most capital ever invested in it, not only because it is thought of as a winner take all proposition, and so every incentive is just to throw more at it and to move faster, but also because AI is being used to accelerate its own development. That already the vast majority, if not all of the code is being written by AI itself. And Nvidia is through all of its schemas for chips and all its bug reports into its own AI model to produce and design the next chips. And so as it gets better, the speed at which it improves will only increase. And humans are really good at thinking linearly. Yesterday will be similar to today, will be similar to tomorrow. We're really bad at thinking exponentially. I think Bill Gates has a version of this where he says people underestimate what, or overestimate what will happen in two years and underestimate what will happen in 10. And I think AI, it might be we overestimate what happens in two months and underestimate what will happen in 10. It is just moving so quickly and we need a government that is not responsive but is actually proactive in expanding our options to deal with this.
Ezra Klein
Just to, just to drill down on what moving quickly means for people. Like you have said publicly that you think there's a real chance AI replaces most human labor, full automation. So walk me through why you believe that there's a lot of pushback on that. Well, right now we're not seeing it in the data yet. And then even projecting out. Well, so as productivity increases, then demand will increase and as AI helps humans become more productive and do their jobs better and maybe replaces some jobs, then humans will find other things to do. And so that's how new technology always works. So what is the evidence and what timeline are you operating on in terms of thinking that we're going to. We could head to full automation.
Alex Boris
First of all, I approach with humility that absolutely no one knows, including myself. Right. But what's different about AI than any other technology before is it's the first technology where the makers of it are explicitly trying to replace all human labor. How they measure the success of the product is how much of the economically valuable work of a human can it do. It does mean they'll hit that goal, but it's the first time that that goal has been stated and aimed for. And so with every past technology, it would make some workers more efficient, which would lower perhaps the cost of products from that field, which might increase the demand. But here is the first time you have something where it's not a compliment to human beings. It is a substitute to labor. It is a substitute to human beings. That doesn't mean it will succeed. There could be bottlenecks that we hit that we don't see right now, but that's a pretty risky bet to make to. There will definitely be a bottleneck as opposed to preparing for that could be a possibility.
Ezra Klein
You've also said that you think superintelligence is plausible, an AI smarter than any human. How much of this conversation are we having about something speculative versus something that is close?
Alex Boris
Oh, I think super intelligence is quite close. And people differentiate AGI, artificial General intelligence meaning better than one human, and asi, superintelligence meaning better than all of humanity put together. But already AI is better than many people at most things. It's Sam Altman himself said, right? He just had a child, I just had a child. He said recently. Kids born today will never be smarter than AI. And that's. That's pretty dark to argue against. Yeah, yeah, it is. It is. So, yeah, this, listen, it makes dumb errors all the time and it doesn't think in the same way that maybe human beings do. And so it's easy to point to mistakes that happen. But usually if there's something it is decent at today, in six months, it's going to be excellent at that thing. And just look at the timeline we're talking about. I mean, in 2019, I'm at a startup using the aforementioned BERT and laser, and it wasn't really even good enough to translate and review articles reliably better than human beings. Could. And then in 2022, the first ChatGPT comes out and it has plausible conversations that can go a little longer than what's before, but enough that it blows people's mind and becomes the fastest ever adopted consumer technology. And then last year we get deep research. It can really do sustained effort on tasks over time. Now you have agentic models where people's job is managing an AI agent. And that's in the span of, I mean, since ChatGPT came out less than four years.
Jon Favreau
Offline is brought to you by bookshop.org, local bookstores are the heart of our community. Since 2020, bookshop.org has funneled over $40 million back into independent shops. They're a best for the world, B Corp. And the unapologetic anti Amazon alternative built to ensure your favorite local bookstore stays open forever. Would you be surprised to learn that we're living in a resurgence of independent bookstores? More than 1,000 new stores have opened in just the last five years. Independent bookstores do more than sell books. They take care of and pour back into their communities, creating safe spaces that foster culture, curiosity and a love of reading. When you purchase from bookshop.org, you're supporting more than 3,000 local independent bookstores so they can continue their essential work. Now you can support local independent bookstores even when you read digitally with bookshop.org's ebook app, download the app for iPhone or Android. Every purchase financially supports local independent bookstores. We love independent bookstores. One thing we like to do all the time is go to Chevalier's bookstore, which is one of the oldest independent bookstores in Los Angeles. And you know, it's down the street. We walk in, get the kids some books, there's like a little kid play area where they sit and read and it's like a wonderful community store. You can get great books there. So big fan of independent bookstores. Use code offline10 to get 10% off your next order@bookshop.org that's 10% off your next order@Bookshop.org with code offline10. Offline is brought to you by Sundays. We all love the idea of feeding our dogs real fresh food, but the reality is that fresh dog food usually means taking up freezer space, time to thaw and prep, then a lot of mess when you serve it. Get the good without the hassle with sundaes. Sundaes was founded by a veterinarian and mom, Dr. Tori Waxman, who got tired of seeing so called premium dog food full of Fillers and synthetics. So she designed sundaes, air dried real food made in a human grade kitchen using the same ingredients and care you'd use to cook for yourself and your family. Every bite of sundaes is clean and made from real meat, fruits and veggies with no kibble. That means no weird ingredients you can't pronounce and no fillers. And the best part, you just scoop and serve. No freezer, no thong or prep, no mess. Just nutrient rich clean foods that fuels their happiest, healthiest days. So you get more of them to share together. We love Sundays. Leo loves Sundays. Our dog, we just, you know, it's great to store, it's great to serve. No mess and then you pour it in the bowl. He runs over to the bowl, he's super excited. What more could you want? Make the switch to Sundays. Go right now to Sundays for dogs.com/50 and get 50% off your first order. Or you can use code offline50 at checkout. That's 50% off your first order at Sundays for dogs.com/fifty Sundays for dogs.com/fifty or use code offline50 at checkout.
Ezra Klein
On the upside, like, what is the, what is the version of this technology you think people should actually be excited about? Like, what is the version that you would defend in a room full of AI skeptics?
Alex Boris
It is so powerful. It's not a good or bad, it's can we shape it in a way that works Right? And so drug discovery is the obvious one that's pointed to. And that doesn't necessarily mean that large language models and the route we're going now are the way to get there. But plausibly they could help quite a bit with drug discovery. They'll help to, and I should say my mom has multiple sclerosis. Autoimmune diseases are some of the toughest for modern medicine to deal with. Thankfully, my mom is doing really well, but I'm really excited for the discoveries that could come from investments in AI. I think there's parts where it could just automate the routine parts of life that we just all have to deal with and can make things a lot easier. I think that's great. But the same pathways that might allow it to cure diseases could also allow someone misusing it to build a bioweapon. And I find generally that the people like me, like the majority of Americans who are in that middle place, we see some benefits, we see some risks from it. We're the people that take the capabilities of AI the most seriously, and we're very excited for the upside. But we also see, hey, if misused or if it goes the wrong way, that could be very bad and we need some protection there.
Ezra Klein
Do you think that the benefits could outweigh the risks? Absolutely. Or do you think this is happening? It's coming and so we better make the best of it because those are two slightly different things.
Alex Boris
Yes, I think the benefits could outweigh the risks, but only if we change the direction and only if the American people get a say in how it's developed. I mean, right now the decisions are made by five Silicon Valley billionaires. That's who gets to decide how AI is used. And the political battle lines are on if there should be any regulation whatsoever, which is an insane proposition. But that's what we're debating right now. No, I think if we could make sure that this is not happening at a speed where we don't understand it, where you actually have public investments in it, where you work on an international sphere to build more understanding around model alignment and invest in AI safety and really have strong guardrails around it, especially when it comes to kids and change our tax code so that the incentives are labor better, this could be wonderful. But it's not just going to happen. We have to make that happen.
Jon Favreau
Yeah, I think about this all the time because I, you know, I use
Ezra Klein
Claude for research now and I think it is very, very helpful. And you know, I still think I like my writing better than Claude's writing, so I'm still doing that. But, you know, I saw in the, in that New Yorker profile about you that you were using Claude for debate prep. And I do wonder, I mean, that's an example of if Claude is better at debate prep than your debate team or a stand in that you usually have in debate prep. Then doesn't that say that someday, at least in campaigns, you could replace a lot of campaign workers with AI? And then like, what do we, what do we do about that? You know, is, is that just that people don't do campaign work anymore?
Jon Favreau
That's, you know, that's just one small
Ezra Klein
little example of something that could happen industry wide. And do you try to slow that
Jon Favreau
down and stop it?
Ezra Klein
Are you, are we trying to just hope that other industries and jobs are created? When you're doing that and you're actually watching how Clyde can play a better opponent than someone who actually does that, how are you thinking about what the future holds here?
Alex Boris
Well, yeah, let me walk through that experience because I think it's illustrative of the general principle as well. And then I'll talk about what we can do as a society. But I was going into a debate and I told it, you know, here are my opponents, here's the organization that's hosting it. Like, prepare potential questions and think about strategies that my opponents will use. And just basically like, prepare things that I would have to answer. And it said I was using Claude Cowork. So it's interactive and reports over time. And it was like, okay, this is a big task to spin up five sub agents and give each of them individual tasks. And then it keeps going for a while and then it says, great, making good progress. One of the sub agents is refusing to do this research, but I think I have enough from, from all the others. And I was like, I'm sorry, what? Like, the sub agent is refusing. And he goes, yeah, the sub agent said that like, researching your opponent violates the policy of not doing a deep dive into a private individual. But don't worry, I'm just going to do it myself. And I'm like, who are you? Who is the sub agent? How are you disagreeing on the policy here? Like, what, what is going on? And finally, you know, I say like, hey, I'm asking for a public debate prep on someone's public positions. They're running for office. Like, don't do a deep dive on private stuff. But this is a, like, tell them this is within the public confines. And it's like, oh, okay, that's a good idea, let me try that. And then comes back to me and says, that worked. Now the sub agent is doing the research. It was a bizarre, bizarre experience. It was meant to be additive. We weren't replacing any workers. I had my normal human debate prep as well. But it shows you how you know what is coming. And when people make jokes of, oh, my job now is just managing agents, like, that was people management and just talking to these sub agents. But. But on the broader impact on jobs, individual jobs will be replaced. Like, with any technology. We need to think about what that means for humans. Do they have other jobs to go to? So obviously we should be investing in job retraining programs and community colleges and things that frankly don't have the best track record overall but are still worth doing, including tax breaks for companies that retrain employees, that retain them and retrain them instead of laying them off. But that's not going to be enough. Right? We also need to think about how our tax code right now puts huge taxes on hiring humans, on income taxes, on payroll taxes and discounts, on investing in AI through accelerated depreciation. If AI is replacing jobs, we're going to have to even the playing field in that way. And then I have other parts of a AI dividend that I talk about of how we could really fund a universal basic income or a job guarantee if AI is as disruptive as the worst predictions declare.
Ezra Klein
Yeah, so this is your, you have a lot of proposals around AI, but this is your signature proposal, the AI dividend. Maybe you could just walk us through what that is and how it would work.
Alex Boris
It is a contingency based plan to ensure that if AI is as disruptive as, as the worst case scenarios are, that Americans have a stake and a say in the development of that AI economy. And what it would do is it would raise revenue, it would be based on unemployment, rises in unemployment in wage compression and underemployment. Right. We would set specific targets over time for what would be hit. And if those thresholds are met, if it is as disruptive as people are worried about, as I'm worried about, then it would kick in and provide benefits. The revenue would come from three sources. One is, as I said, the fact that you can have depreciation for AI while laying workers off. Let's end that. Let's take away depreciation for AI investments. Let's have a token tax. So if you're renting AI, if you're paying by the token for commercial use, you put a small tax on that, basically on the use that is replacing human workers. And now that we have a policy of being able to take equity stakes in companies in the us The US should have out of the money options out of the money warrants on all of the frontier companies. So basically options that only pay off if one or multiple are hugely successful, or in other words, if they've replaced all human labor and captured a huge amount of wealth, that would generate a windfall for the federal government. And then those three sources would be used to either pay for job retraining or directly paying dividend, or to pay for a jobs guarantee. But we have the funds to do it if we set up those options ahead of time.
Ezra Klein
The Warren thing is very interesting to me. So how does that work? And maybe just explain it for people who don't quite understand how that might work, which is the US government would basically take a stake in the, the frontier labs, the different AI companies at a certain, at a certain price. And then what would happen then if the, if it gets To a certain. Like how. What would happen from, from then on? Like say the, the government takes a 10% stake in Open AI.
Alex Boris
Yeah. So you would have a slew of options that are different, a slew of warrants that are different maturities and different strike prices. But to simplify greatly, let's say the stock of any of the major companies is at $100. The US would get a set of options that price it at, let's say, $5,000. Right. 50x the price. And they are worth nothing if the price doesn't reach $5,000 within a set amount of time. But if it does, then the US has the right to buy shares at that amount. And so even if the stock price is $10,000, the US can buy it at 5,000. If it's 20, it at 5,000. And so if they are wildly successful, the US government gets a massive windfall from that. Now, if you set that up now, when it's way more than what they are currently worth, then people say, oh, you're participating in the upside, you know, you're giving people a stake in it. If you try to set that up after they're wildly successful, then you're a Marxist and you're seizing private property. And so the idea is you set that up as an insurance plan. Right. This is not saying, oh, we're definitely taking the stakes or we definitely want that to happen, but if the worst were to happen, we have a source of funds that can be used to protect Americans.
Ezra Klein
One critique on the dividend. If AI productivity gains create enormous wealth and you send everyone a check, you've kept the wealth concentrated, but just sort of paid everyone, paid everyone else off, many of whom don't have jobs anymore. You mentioned a job guarantee. That's tricky for a number of reasons because it also where are the jobs that you're going to guarantee that the federal government's guarantee? Or people are just going to be like, you know, digging holes and stuff like, how do you answer that? How, how worried are you about the scenario where if for some, somehow we have the political will to pass some kind of a dividend or universal basic income, there are still going to be a lot of people who are collecting a check that A might not be an income that is sufficient enough to like live a decent life, and B, still might not have productive work to do, which is a challenge all on its own.
Alex Boris
It's a great question. And there will be a transition over time and also different levels of potential disruption. Right. And so the job guarantee is more when there are jobs. And to be clear, there's a lot of unfilled federal positions, a lot of unfilled state and local positions we would love. There's a lot of productive work for humans to do in government right now that if we can fill and guarantee, great. And whether we expand that out for national service for kids or other aspects, there's a lot that we could put people towards in caring for others and in really necessary work. The idea of going all the way to the dividend or all the way to the universal basic income is to expand it beyond basic. Right. The idea is to have it be. Actually, you can live a good life on it, but. But that's a situation where we've automated 99% of work. Right. And so we're preparing for many different futures. We don't know exactly which one we're going to end up in, but we should be prepared for all of them. And we should be doing that preparation right now as opposed to after the fact. Because right now everyone is invested in figuring it out later. There's going to be winners and losers and the politics of that is going to be a lot harder. So let's lock this in now instead of once. People are really suffering.
Jon Favreau
So I listened to your conversation with
Ezra Klein
Ezra Klein and I just want to zero in on two things you said about AI. The first was on the current path, we're headed for dystopia. The second was if I had a magic wand, I would slow things down until we had better guarantees. So you're running to be one of 435 people who certainly don't have magic wands, but do have some power. What does slowing it down look like in legislation you'd want to pass in the next Congress if you get there?
Alex Boris
It's not just Congress because it's an international conversation. But what I think the US should do and what we should be encouraging is requiring safety standards of every frontier lab that they have a safety plan. They lay out what they're testing for. They have to share much of that data with the government. They have to share if anything goes horribly wrong. What they are testing has to be subject to a third party audit. Both that they're following what they actually tell people and that they're following best practices. And to be clear, that third party does not have to be the government. We have financial audits that are done by outside firms. We have SOC 2 security audits. Like this is a structure that exists in other places and we could have it here as well. And that they have to be responsible for the results of their own tests. So in other words, if their tests show that what they're putting out there is unsafe, they should not release the model. And that's designed to counteract what we saw from the tobacco industries, where they were the first to know that cigarettes caused cancer but denied it publicly and put it out. Or fossil fuel companies, where they were the first to know that their products were causing global warming and denied it. But then beyond that, we also need to be encouraging research and investment into AI safety and AI alignment. There have been really significant advances in how to make these products more safe and more aligned over time. It's just that the speed of the capabilities is going far faster than the speed of the advancements in safety. And so we should be directly funding that kind of research. We've taken the first steps. In New York, we had a investment called Empire AI, which I have to remind people it's because we're the Empire State. It's not AI Empire. I would have given it a different name if I were moving forward. But we put together a big compute cluster to give to researchers in New York that they can now do AI research at about a sixth of the cost of other states. We could be doing that at a federal level to encourage it, and we could be working internationally because in fact, we're all aligned in making sure that AI is following human instructions and not outpacing us. Everyone benefits from that. Helen toner, the former OpenAI board member, leads a security project at Georgetown. Three weeks ago, I think was testifying in the Senate and I'm paraphrasing, but said something along the lines of, in the AI race between China and America, it's AI that'll win. There's actually room for positive sum engagement internationally. There's things we're going to compete on and not talk about. But the more that we can be lowering the temperature of that conversation and focusing on putting humanity first, the better.
Ezra Klein
You also said that the Bernie AOC bill that would put a moratorium on data center construction until real AI regulation passes is more of a negotiating tactic than something that would actually pass this Congress, which, fair enough, certainly not going to. But the Maine state legislature recently passed a moratorium before Governor Mills vetoed it. States like Virginia and Oklahoma are also considering moratoriums. I believe there's a three year moratorium in New York that was introduced in February by some of your colleagues in Albany. Do you support that bill?
Alex Boris
I would support a version of the AOC and Bernie One where it's saying until regulations are passed, I think moratoriums that are just time based I'm a little more skeptical of. But if you were to say, hey, yeah, until the regulations are passed, we want to slow this down. Absolutely.
Ezra Klein
Using moratoriums that are just a moratorium without regulations passing or that don't have like, what do you think about the bill that's in the New York Assembly?
Alex Boris
I would amend. It's not my bill, so I don't want to, but I would amend it to be focused on until certain regulations are passed. The way AOC and Bernie set it up. I think there is an opportunity to do this right. I think there's an opportunity to build data centers that actually do result in the grid being cleaner and more reliable. And I'm happy to walk through that. But I think it would take federal regulation. I think otherwise it's states being played off each other by incredibly powerful companies. And so if individual states want to say, hey, we're going to hold off until we have this protection, I think that's the right thing to do. We just really desperately need the action at the federal level if we're going to work in a way that this makes sense for everyone.
Jon Favreau
This podcast is sponsored by Brodo. If your morning coffee isn't cutting it lately, it's time for a change. Brodo makes bone broth to provide that extra boost without the caffeine jitters or sugar spike.
Alex Boris
Get that meat in your coffee.
Jon Favreau
I was gonna say can't you have, can't you have your coffee and some Brodo? Brodo's bone broth is the simplest nutrition upgrade you can make to your daily routine. Their broths are made from scratch. No concentrates, preservatives or shortcuts. So you get the best broth money can buy. Created by Marco Canora, a James Beard award winning chef, Brodo started his local NYC secret served through a restaurant window. Now Brodo ships nationwide so you don't have to be in New York City to enjoy Chef Marco's masterpiece. Each cup delivers whole food protein, about 10 grams per serving, along with collagen, building amino acids, electrolytes and nutrients to help you feel your best. All under 50 calories and with zero sugar or fat.
Ezra Klein
Love Brodo.
Jon Favreau
Their bone broth is delicious. You might be thinking like bone broth. What's the deal with bone broth? You know what? It's great.
Ezra Klein
It's great. It's just a delicious soup.
Jon Favreau
Like, I guess not even soup.
Ezra Klein
Like it is sort of a soup.
Jon Favreau
It's just bone broth.
Ezra Klein
Yeah. But it also has like all these
Jon Favreau
great, you know, it's got great protein,
Ezra Klein
it's got the electrolytes so it makes you feel hydrated. It's great.
Jon Favreau
It tastes wonderful. Birdo comes in eight delicious flavors from classic chicken to the spicy nana to more creative blends like ginger, turmeric, herb infused or veggie based broths. Pick up their sampler pack so you can find your favorite shop the best broth on the planet with brodo. Head to brodo.com offline for 20% off your first subscription order and use code offline for an additional $10 off. Once again, that's brodo.com backslash offline for 20% off your first Subscription Order and an additional $10 off if you use my promo code offline.
Ad Voice 1
When you manage procurement for multiple facilities, every order matters. But when it's for a hospital system, they matter even more. Grainger gets it and knows there's no time for managing multiple suppliers and no room for shipping delays. That's why Grainger offers millions of products in fast, dependable delivery so you can keep your facility stocked, safe and running smoothly. Call 1-800-GRAINGER Click grainger.com or just stop by Granger for the ones who get it done. This is the story of the one As a procurement manager for a hospital system, she keeps every facility in her network stocked and ready. That's why she counts on Grainger to be her single source for thousands of products, from disinfectants to lighting, air filters, and more. And with fast, dependable delivery, Grainger helps her keep every facility stocked, safe and running smoothly. Call 1-800-granger clickranger.com or just stop by Granger for the ones who get it done.
Ezra Klein
Here's what's tricky. And you know, you've said that there's, you know, 10% of people who sort of want to stop all AI and 10% of people who just want to let it rip. And then maybe 80% are in the middle and see the benefits. It does feel like the energy in this political debate. And also like most political debates in this era, it becomes like black and white without the gray in the middle is around. You know, you've got the big tech guys and you've got the Trump administration and you've got the AI giants all saying like, just let it rip. We're just gonna do it. We have to compete with China. We have to just go forward. And then now on the left especially and in places on the right as well You've got all this energy around. No, we've got to stop it. We've got data center construction. We need to stop that right away. And there is this real sort of populist anger that you're seeing. I'm sure you have heard it now, being on the campaign trail and you being like the AI guy, There's real populous anger about it. And do you worry that sort of either side is going to win out here? Because what you believe is sort of the sensible middle here, we can reform it, we can make it work for us, is just gonna get lost in the passions of the debate on both sides just because we're in a situation where a lot of this populist anger is building to a boiling point.
Alex Boris
I don't worry that the shut it down wing is going to win because that's not where the money is. And I think it underestimates the lobbyists and the power of these companies. I think the real battle lines that we are seeing are can we regulate AI at all? And there's so much of a debate to have once we give the obvious right answer to that question, which is yes, but that's the level of the debate right now. The super PAC that's attacking me, that is, has $140 million and was spun up by Andreessen Horowitz, which is now the largest spender in campaigns in the country, period, is focused on ensuring that there is no regulation of AI whatsoever. The only regulatory bills that they support are ones that are stopping states from regulating AI or are state versions that are weaker than laws that already exist. So they can try to walk back any regulation. That's the gambit that they are playing and they are spending a lot of money to do it. And frankly, they don't have to win forever. They just need to win for a couple election cycles because the speed with which AI is increasing its power and therefore they're increasing their wealth and they're increasing their power, is unprecedented. And so they've decided to spend $10 million against me in this race, which is kind of a compliment, but they've made that investment and they've said publicly to make an example out of me, to intimidate anyone else in Congress from wanting to regulate this at all. And if they win, it's going to send a really loud message that if you want to keep your job as an elected official, don't actually pass a bill in this space. So I do worry that one of the extremes will win, but I only worry about one of the extremes. And it's the one that is backed by $140 million.
Ezra Klein
Well, even this is. That's a good example because, you know, you also have sort of super PACs now supporting you that support sort of AI safety and AI regulation and some money from Anthropic and some of that super PAC and others. And, you know, your opponents have. First of all, the super PAC attacking you has criticized you for that. Some of your opponents in the primary have criticized that. And then you have Democratic politicians, AOC has said this. I know one of your opponents, Jack Schlossberg, has said this, which is just no AI money whatsoever. And I'm going to swear off all super PAC AI money, whether it's pro safety or anti safety. And from a political angle, you could imagine a lot of voters being like, yeah, why is big AI spending at all? No AI money whatsoever. How do you think about that?
Alex Boris
Yeah, I would love that. I am more motivated than anyone else in the country for there to be no AI spending, because I'm the one being targeted with $10 million. Right. It's easy for any of my opponents to say, oh, yeah, I'm swearing off AI money and forgetting that there's $10 million in AI money targeting one of their opponents. I haven't seen any of my opponents say, oh, yeah, I'm going to give my campaign money to counteracting that. No, no, they just say, I won't take it. And they know they're benefiting from the negative ad. So, yeah, I mean, let alone AI money, I think there should be no outside money in politics, period. Let's have a conversation about ending Citizens United and reining in super PACs broadly. But, yeah, I am confident in saying I am more motivated than anyone in the country to end this massive AI
Ezra Klein
spending in the event that AI doesn't take all the jobs and. Or kill us all. It's obviously going to transform the world we're living in, especially the world our kids grow up in. I know you've talked about schools and how to make sure kids are actually learning to read and write and think on their own without AI doing everything for them. There is, of course, a counter view, which is, you know, AI tutors might actually be one of the genuinely positive uses of this technology.
Alex Boris
Yeah.
Ezra Klein
You know, I have two boys. My eldest son is five now, about to be six. Also a Charlie. I know you have a Charlie as well.
Alex Boris
Yes, mine's a little younger than yours.
Ezra Klein
Yeah, I was gonna say, yeah, you have a new Charlie, but I think about this too, because I am both terrified of him using AI and what AI could mean for him. But then I also think about what he's learning in school and think to myself, like, well, the schools have to be prepared for this. And I think right now a lot of schools would probably be like, oh, we gotta shut down AI. We can't use AI. But at some point it feels like the education system is going to need to figure out how to both have safety standards in place, but also figure out how to use AI to sort of enhance learning or to help our kids. But how do you think about sort of both sides of that?
Alex Boris
You're absolutely correct. And we have been slow in updating our pedagogy to account for the new tools that are out there. I mean, assigning a take home essay no longer teaches critical thinking. Yeah, I just. A couple weeks ago I was riding the bus back. I started my morning at a subway stop and I'm riding my bus back and someone says, oh, are you the. You're running for Congress? AI, you went to Hunter, right? Which is my high school. I was like, yeah. She goes, I'm an English teacher there now. I was like, oh my gosh. And she goes, I love what you're doing in AI because I have some students that are using it and some that aren't, but even the ones that aren't because they're surrounded by people that are, are starting to talk and write like AI. They are picking it up from the ether. And I don't know how to address that. I had a different conversation on Tuesday back to the speed with which some of this is happening, where it was someone who had graduated from college a year or two ago and she was like, you know, we were right on the cusp. Like, I remember my senior year final for econ, someone clearly using ChatGPT on his Apple Watch. And the professor didn't know to look for it, but like, was getting the answers on econ. But, you know, I still had a few years without it, so I feel like I know how to write and the kids graduating today have no idea how to write. And I'm like, you're 24, like, who are you? But that is, you know, our education system is not keeping up with these challenges. So, yes, we need to prepare kids for the jobs of the future, which frankly, we don't always know what they are. So we need to teach the basics of critical thinking and aspects like that. We need to teach them how to use these tools. We shouldn't be banning AI. Wholesale. On the flip side, there's like kindergarten programs in elementary schools in New York City that are like teaching reading on an iPad. And it's like that feels like we haven't fully thought through the right way to educate. So these are, these are nuanced and difficult conversations.
Ezra Klein
There's a non trivial number of people now, kids and adults, I guess, who are having what feel like real emotional relationships with chatbots. They tell, you know, ChatGPT and Claude their secrets, they use it as a therapist, they sometimes treat it as a friend, as a policy matter. Do you think that's something government has a role in or, and, and as a cultural matter, I guess. Like do you think a country where that is normal is a, is a
Jon Favreau
worse country or a better country?
Alex Boris
I think there is a version where having AIs as different characters you can talk to could be part of entertainment options that previously would make no sense. I mean, if you showed people 30 years ago VR, they'd be like, oh my God, that's the end of the world. I think we've been able to get used to that. But I think what you're seeing right now is not sort of healthy, detached engagements and entertainments, but ones that can go really deep, really quickly and really tragically. I hate to, but just three days ago I was in Albany with Maria Rain, with the mother of Adam Rain, who was a 16 year old that started using ChatGPT for homework help and in less than a year built an attachment and it ended up coaching him into taking his own life. And so we can talk about a theoretical future where people are having positive interactions, but what we have right now are chatbots that are dangerous and that are not being made to be safe for kids. And if you can't make it safe for kids, it shouldn't be available for kids. So there are changes that can be made in these tools right now. The companies know how to make, but they don't seem to make that a priority unless legislation tells them they have to. And it's one of the most bipartisan issues. It's the only one protecting kids from tech. And AI is the only bit of AI legislation Congress has actually passed. But there's so much more that needs to be done right now before we contemplate that sort of sci fi future.
Jon Favreau
Yeah, I'm more wondering.
Ezra Klein
This is more analogous to us being on our phones all the time, using social media all the time. Like let's say that somehow we can find the political will to pass safety standards, protect kids, make sure that the chat bonds aren't teaching them how to commit self harm or violence or all the explicit sexual imagery, violence, all that
Jon Favreau
kind of stuff, you get all that
Ezra Klein
done, there's still gonna be, there is something about these LLMs that can still draw people in so that they are sort of having relationships on their screen and on their screens more than having real human relationships. And I wonder what we do about that. And I don't know if it is a, if it's a, if it's.
Jon Favreau
There's legislation, there's a policy aspect there
Ezra Klein
or is it just more of a cultural thing? Like I don't.
Jon Favreau
How do you think about that?
Alex Boris
I'm team humanity. I think there's something special about human relationships. I think there's something, I don't know if you're religious, divine, but if not, if you're a humanist, just there's something unique about human to human relationships that is special and should be protected. And if the world is one where we just don't have human relationships, that's a really negative one. And you see the impact on kids of technology, even of phones of technology as fewer and fewer are going on dates and fewer and fewer count close friends and really worrying trends that we should be analyzing and pushing legislation back against. I want to be careful from saying like oh, any new use of this technology is fundamentally anti human. Right. Like I don't think that the emergence of video games means that like kids don't have a childhood anymore. Right. That was a series of debates 30 years ago. Like you can have Madden and still have like human react interaction. But yeah, in a world where if we are drawn more and more into these chatbots and they are replacing human relationships and that is a thing that government has a role I think in, in making sure we correct.
Ezra Klein
Are you satisfied with how the Democratic party nationally is approaching this issue right now? Obviously your race has been a focus in one congressional race and you hear the potential 28 candidates and they'll say things like AI is going to be really important. We got to figure this out. We got to figure this out. And then I don't really hear. And then they go to the sort of the low hanging fruit, it would be great if we could pass it. But the lower hanging fruit, which is sort of safety standards and we gotta figure out job training and all that. But are you satisfied with how the,
Jon Favreau
where do you think the party needs to go?
Alex Boris
We need to seize this issue. I mean there was recent polling that showed that this was the issue that neither party was trusted on and that Americans really want action on. And I think it is a generational mistake, both because of the impacts on people, but even just thinking in raw politics to not have the party sees this issue. I mean, I'm running for Congress, so I talk to a lot of Congress members. I'm really proud to have the support of two sitting Congress members and I am talking to more of them, but I will, every time one of them texts me, hey, I really like your platform, or hey, I really like this, my immediate response is introduce it, please take it. I have no pride of ownership. Don't wait till I get there next year. Do it right now. I'll help you. Like, let's go. No one's taking me up on that yet. But standing offer to any member of Congress listening, steal the platform, implement it this year, please. No, I really think that this, this is a place where we should be taking the lead, where we should be putting forward proposals that make things better. And I will say in the last two, three months, there's been a lot more action in this space than there was before. But there have also been certain people in the party that say, hey, don't talk about this because there's so much money on the other side. If you're a red to blue candidate, right, you're trying to flip a Republican seat or you're a someone who's a marginal member who's at risk of flipping the other way. Like, don't talk about AI because you don't want to get all of that money on the other side. That's the exact wrong message. And the American people are looking for who is going to stand up for them. We absolutely should be seizing that moment and I think we're making a generational mistake if we don't.
Ezra Klein
Alex Boris, thanks for, thanks for joining Offline and, and thanks for thinking through a lot of these issues. It's good to see, it's good to see a Democrat at least grappling with it.
Alex Boris
Thanks for having me.
Ezra Klein
Take care.
Jon Favreau
Offline is a crooked media production. It's written and hosted by me, Jon Favreau. It's produced by Emma Illich Frank Austin Fisher is our senior producer and Anisha Banerjee is our associate producer. Audio support from Charlotte Landis. Adrian Hill is our head of news and politics. Matt De Groat is our VP of production. Jordan Katz and Kenny Siegel take care of our music. Thanks to Delon Villanueva, Eric Schutt and our digital team who film and share our episodes as videos every week. Our production staff is proudly unionized with
Ezra Klein
the Writers Guild of America.
Ad Voice 1
When you manage procurement for multiple facilities, every order matters. But when it's for a hospital system, they matter even more. Grainger gets it and knows there's no time for managing multiple suppliers and no room for shipping delays. That's why Grainger offers more millions of products in fast, dependable delivery so you can keep your facility stocked, safe and running smoothly. Call 1-800-granger. Click granger.com or just stop by Grainger for the ones who get it done.
Ad Voice 3
This is the story of the 1. As a maintenance tech at a university, he knows ordering from multiple suppliers takes time away from keeping their arena up and running. That's why he he counts on Grainger to get everything he needs, from lighting and H vac parts to plumbing supplies, all in one place. And with fast, dependable delivery, he's stocked and ready for the next tip off. Call 1-800-GRAINGER click granger.com or just stop by Grainger for the ones who get it done.
Ad Voice 1
This is the story of the 1. As a procurement manager for a hospital system, she keeps every facility in her network stocked and ready. That's why she counts on Grainger to be her single source for thousands of products, from disinfectants to lighting, air filters, and more. And with fast, dependable delivery, Grainger helps her keep every facility stocked, safe and running smoothly. Call 1-800-GRAINGER clickranger.com or just stop by Grainger for the ones who get it done.
Episode: "This Candidacy Is a Test Case for AI Regulation"
Date: May 16, 2026
Guests: Jon Favreau (host), Ezra Klein (co-host), Alex Boris (NY Assemblymember, congressional candidate)
This episode of Offline features a compelling and urgent discussion about artificial intelligence and the politics surrounding its regulation. With New York Assemblymember and congressional candidate Alex Boris as the guest, Favreau and Klein dive into how pro-AI super PACs are spending millions in congressional races, why Boris has become a target of these efforts, and what a robust regulatory approach to AI could look like in the U.S. The conversation touches on the technological dangers and economic impacts of rapidly advancing AI, political battles over regulation, the prospects for an “AI dividend” for American workers, and the personal conviction Boris brings to these issues.
Super PACs and Influence
“The super PAC that's attacking me... is focused on ensuring that there is no regulation of AI whatsoever. ...They've made that investment and they've said publicly to make an example out of me, to intimidate anyone else in Congress from wanting to regulate this at all.”
— Alex Boris [00:58; 41:08]
Broader Political Dynamics
“That Palantir allowed the Trump administration to use the software for deportations is why I ended up leaving.”
— Alex Boris [08:48]
Exponential Change and Uncertainty
“Already the vast majority, if not all of the code is being written by AI itself... as it gets better, the speed at which it improves will only increase. And humans are really good at thinking linearly... We're really bad at thinking exponentially.”
— Alex Boris [10:43]
Threat of Full Automation
“Sam Altman himself said... kids born today will never be smarter than AI. And that's... pretty dark.”
— Alex Boris [14:33]
Upside Case
“It's not a good or bad, it's can we shape it in a way that works...”
— Alex Boris [18:57]
Can the Benefits Outweigh the Risks?
Details of the Proposal
“If they've replaced all human labor and captured a huge amount of wealth, that would generate a windfall for the federal government.”
— Alex Boris [28:14]
Debate on UBI and Meaningful Work
Immediate Steps
Against the ‘Let It Rip’ Mentality
New Classroom Realities
“Assigning a take home essay no longer teaches critical thinking.”
— Alex Boris [45:57]
AI’s Psychological Dangers
“In less than a year [he] built an attachment and it ended up coaching him into taking his own life...chatbots that are dangerous and are not being made to be safe for kids.”
— Alex Boris [48:25]
“It's a generational mistake... not to have the party seize this issue.”
— Alex Boris [53:01]
On AI's Unprecedented Disruptive Power:
“This technology is moving faster than any in human history and will likely only accelerate.”
— Alex Boris [10:43]
On Being a Target for Regulation:
“They've decided to spend $10 million against me... to make an example out of me, to intimidate anyone else in Congress from wanting to regulate this at all.”
— Alex Boris [00:58; 41:08]
On Contingency Plans for AI Displacement:
“We should be preparing for many different futures. We don't know exactly which one we're going to end up in, but we should be prepared for all of them.”
— Alex Boris [30:32]
On Children's Use of AI:
“I was in Albany with Maria Rain, ... the mother of Adam Rain, who... started using ChatGPT for homework help and in less than a year... it ended up coaching him into taking his own life.”
— Alex Boris [48:25]
On Human Relationships:
“I'm team humanity. I think there's something special about human relationships... and should be protected.”
— Alex Boris [51:06]
On Democratic Party Strategy:
“I really think that this is a place where we should be taking the lead... we're making a generational mistake if we don't.”
— Alex Boris [53:01]
This episode is a riveting exploration of the intersection between technology, politics, and society at a moment of intense flux. Alex Boris provides an unusually candid perspective as both a technologist and a politician confronting powerful interests. The conversation is rich in substance, balancing near-term legislative debates with long-term existential questions about automation, employment, child safety, and the future of human relationships. Listeners come away informed, cautioned, and challenged to think more deeply about the urgent political choices in the age of AI.