Money Talks: Is it Time To Panic About A.I. Yet?

0:00Slate Money is brought to you by Charles Schwab. Decisions made in Washington can affect your portfolio every day. But what policy changes should investors be watching? Listen to Washington Wise, an original podcast for investors from Charles Schwab, to hear the stories making news in Washington right now. Host Mike Townsend, Charles Schwab's Managing Director for Legislative and Regulatory Affairs, takes a nonpartisan look at the stories that matter most to investors, including policy initiatives for retirement savings, taxes, and trade, inflation concerns, the Federal Reserve, and how regulatory developments can affect companies, sectors, and even the entire market.

0:41Mike and his guests offer their perspective on how policy changes could affect what you do with your portfolio. Download the latest episode and follow at schwab.com slash Washington Wise or wherever you listen. Slate Money is sponsored this week by Your Next Dollar, a podcast from NerdWallet Wealth Partners. Your Next Dollar is built specifically for high earners like you. It covers the financial complexity that comes with a higher income. If you have equity decompensation, RSUs, tax planning, housing decisions, figuring out when it's actually okay to enjoy your money? It turns out that the more you earn, the more complicated the questions get.

1:21And yet, somehow, there's even less content built for you until now. Your next dollar covers both sides of the equation, building wealth and spending on what actually matters to you. Because at a certain income level, the question isn't, should I save more? It's, what should I do with everything I have. Financial advisors from NerdWallet Wealth Partners join the show regularly to discuss the questions they hear from high earners like you. The nuanced, real-world stuff generic financial content doesn't cover. If you want help getting more out of every dollar you earn, follow Your Next Dollar wherever you get your podcasts.

1:59Sponsored by Your Next Dollar, a podcast from NerdWallet Wealth Partners, an SEC-registered investment advisor. This podcast is for informational purposes only and is not investment, tax, or legal advice.

2:17You live in the Fed. Yes, absolutely.

2:23Hello and welcome to Money Talks, a special extra podcast from Slate Money where we chat with brilliant and interesting people. I'm Emily Peck. I write the markets newsletter at Axios and co-host Slate Money. Lately on the show and in the newsletter I write at Axios, all we seem to do is talk about AI. From our perspective, it's all about investment, right? Investment in AI is driving the economy. It's driving the stock market. But also it's doing something else, which is, to put it in a very technical way, it's scaring the hell out of people. Earlier this year, everyone sort of woke up to this notion that AI could maybe go rogue.

3:04I don't know. Kill us all. After the Hugging Face incident, that's when OpenAI revealed that a group of AI agents had broken containment and hacked into the computer systems of another company, Hugging Face. And since then, there's been a drumbeat of sort of similar news leaks where open air Anthropics says, oh, our agents did this or that. And there's been increasingly loud calls to regulate AI like we've done in the past for other scary technologies, cars, nuclear weapons. Yes, cars. I count as a scary technology. Anyway, I wanted to talk about this with the people on the ground kind of watching this all unfold.

3:47And for me, that had to be just two amazing reporters from Axios, where I work. And they're both going to come in and talk to me for this episode. First, I'm going to talk to Sam Sabin. She's a cybersecurity reporter at Axios who's going to join and talk about sort of the real threats when it comes to AI, what she's seeing, what's alarmist, what's real. Then after a break, we'll get to Maria Curie. and she covers tech policy and has basically been on top of the Trump administration's every move in the space. And she'll join to talk about how the White House and lawmakers at the state level in particular are handling this fast moving situation.

4:31It's pretty fun. So stick around. We'll be right back.

4:42Today's episode of Slate Money is brought to you by Vanguard. To all the financial advisors listening, let's talk bonds for a minute. Capturing value in fixed income is not easy. Bond markets are massive, murky, and let's be real, Lots of funds, throw a couple flashy funds your way and call it a day, but not Vanguard. Vanguard bonds are institutional quality. It's a commitment to your clients. It's top-grade products across the board. There's more than 80 bond funds. They're actively managed by a 200-person global squad of sector specialists, analysts, and traders, of course. This isn't about star portfolio managers.

5:25It's about active strategies that aren't locked away with one person, that is shared across the team so that every client benefits from the collective brainpower. No one individual can really mentally encompass the bond market. It's way too complex. So if you're looking to give your clients consistent results year in and year out, go see the record for yourself at Vanguard.com slash audio. That's Vanguard.com slash audio. All investing is subject to risk. Vanguard Marketing Corporation Distributor. This holiday season, don't miss your chance to turn just looking into just bought it. Make it easy for your customers to say yes on the spot.

6:05PayPal Checkout is a trusted conversion engine fueled by 400 million active PayPal and Venmo accounts. And when you add Pay Later, you can turn holiday shoppers into regulars. 96 % of Pay Later shoppers are repeat users. So, if you give them pay later flexibility now, it could help earn their loyalty all year long. PayPal is built for all business. Get started at paypal.com slash business. PayPal data, 8 markets, March 2025 to February 2026. Repeat users defined as more than one pay later purchase in the past year.

6:46Every restaurant wants to be busy. Most restaurants aren't ready for it. I'm Chef Mike from Detroit 75, and Toast helps us stay ready when the lunch rush hits. We prepare for lunch like we're preparing for battle. Orders are flying, tickets are stacking, and the guests should never fill the chaos. Toast keeps our team organized so we can serve more people, stay consistent, and keep growing. Visit toasttab.com to learn more. Pay testimonial from an actual Toast customer.

7:14Welcome back to Money Talks, and I'm joined now, as promised, by Sam Saban. She is an incredible cybersecurity reporter at Axios. Sam, welcome to Money Talks. Thanks for having me, Emily. I'm so excited. I'm so excited. So you cover cybersecurity and hacking and have done so even before AI was all anyone was talking about. You were like on top of the whole situation. And it's already kind of a scary beat, I can imagine. I've heard you talk, I think, at our latest retreat, and you just sounded so paranoid about everything AI-related. We were being told to use it, and you were like, but there are these issues.

7:56While everyone's sort of worried about AI and hacking, you've been reporting the AI hacking crisis is already – it's basically happening already. What has changed, say, in the past six months? Has it made you more worried? What does the AI hacking crisis look like right now to you? So interesting that whenever someone says I come across as paranoid because I'd argue that compared to the people I talk to, I'm more measured. I'm not as scared. I think in the last six months, two things have happened that have been a little bit different in terms of AI and its impact on cybersecurity. One, the capabilities have changed.

8:39I'm sure people have seen all the headlines dating back to April, which was only six months ago, which feels like a lifetime ago for whatever reason in this world, when Anthropic started warning people about a scary new model called Mythos that would be able to find bugs and zero days and throwing around all these words that really, I think, lit up a lot of people across enterprises who maybe had been taking cybersecurity for granted or didn't think that the AI models were there. there. And then two, I think as a result of that, there's been a lot of attention paid to cybersecurity. And you have a lot of people who are new, who maybe are not in the industry, or it was maybe a secondary thing they had to worry about on their checklist that are now starting to learn about it.

9:27And so you're getting a lot of research and headlines that are filled with fear, uncertainty, and doubt, or FUD, a fun acronym that we like to use in the industry. sometimes we have fun here, that I think is really, really, really turning up the temperature around the concerns around cybersecurity threats. And so things are scary. Models are definitely capable. We're seeing more and more cyber criminals using AI in their attacks. But at the same time, you're also getting a very alarmist, the sky is going to fall down, the internet's going to come crumbling apart and you can't access your money or whatever as well.

10:11That's really making it hard to figure out what's real, what's not. And it's really difficult to parse through and I live in this space day in and day out. You live in the FUD. Yes, absolutely. I think I was reading one of your stories and it was like, hackers can now buy AI hacks online really easily. What is currently the threats that companies enterprises, you call them? What are they actually dealing with? It seems like a lot of the usual kind of things, but maybe I'm wrong. No, you're right. And that is kind of the interesting part is that we're thinking, oh my God, things are so scary. I guess I'll say this in a way to make it sound more positive, even though it's still bad, which is things were scary before.

11:01They're just now a touch scarier because ai you know although the hacking techniques we're seeing are still the same more act threat actors who maybe were not as skilled can now use ai to be more skilled and sophisticated and level up or they can launch more attacks right so think about a scammer who was charged with sending Emily text being like, you know, I see that you haven't paid your toll recently on EasyPass. Click this link. What they can now do with AI, whereas before maybe they were just kind of using some tools to go through a pre-bought list of phone numbers that may or may not be active and then sending Robotex that way.

11:53What they can now do is use AI to kind of mine for data that is even more specific. So rather than texting you about EasyPass and you go, I don't remember the last time I went through EasyPass or I have an automatic payment or however it works, it can be more specific now, right? So instead they could pose as maybe a job recruiter who wants to hire you to work at a news publication and they're offering you more money or more senior title and you're like oh wait this sounds amazing and it's because i know right you should really figure that out uh but you know they can then mine your information so much faster because they're using llms that are maybe reading through your linkedin that are reading through all the other information that was stolen about us and leaked on the dark web for eons and eons.

12:47We all get these alerts all the time. And it's just a lot faster. And when it comes to this is basically I'm talking about social engineering, right? Tricking people into falling for scams. That gets way easier. And then you can add it to enterprises and so on and so forth. And that's just one tip of the iceberg. I haven't even gotten to software bugs and how much easier it is to find those and companies or zero days. We're talking about technology that has made it so that the same hacking techniques we've always heard about that maybe have led to the shutdown of a major gas pipeline, if people remember like five years ago when that happened around Memorial Day and then people were putting gas in plastic bags.

13:29That was crazy, by the way, and wild. We were talking about the ability to do similar incidents, but at a scale and a speed that most companies are not ready for. I think a lot of companies spent a lot of time thinking, well, I won't be hit or I won't be the target. Who would think about little of me? And now it's just you have no idea what an AI agent is going to be able to find or why a threat actor would go after you or anything like that. It's just there's a lot of unknown. And I think there are a lot of weak points in our online systems that people just didn't think to fix because it's kind of similar to going to the dentist.

14:10It sounds annoying. It sounds expensive. It sounds painful. And now we're at the point where perhaps some organizations are going to need like root canals or implants or they have to do it fast, right? And be willing to do that really, really fast before threat actors are willing to scale. So that sounds like, I mean, at the same time, companies are panicking that all their weaknesses are about to be exploited. There's got to be a flip side to this. If AI makes it easier to hack, doesn't it have to make it easier to prevent against hacking? And is that also happening on the other side of the war or whatever this is?

14:47The idea of a war is very cyber. Yes, absolutely. I think what has been so fascinating over the years of covering this is that even though AI models have gotten a lot better, if anyone listening uses them, I use them all the time. It is astonishing what Claude or ChatGPT can do for me day in and day out. The advice that I hear from security professionals hasn't changed, right? They are saying two things. One, focus on the basics. If you don't have multi-factor authentication, which is, you know, when you probably get that text that's saying, oh, here's the login pin for whatever you insert a password.

15:29If you don't have that at your company yet, perhaps do that. And then two, use AI. Use AI before the bad guys are using AI on your systems, right? Use it to find the bugs, use it to figure out how to patch them, and then use it to help test out those defenses. And the only challenge there is that unlike hackers who they don't have HR, they don't really have to worry about, you know, sales or working with vendors to get enterprise software that can help them with this. They're just kind of going and they're, you know, throwing things at the wall and whatever sticks they're doing. You know, unlike hackers who maybe are more free to do whatever they want, enterprises are stuck in procurement and like legal and compliance and they got to train their employees.

16:22They have to figure out which vendors are working. And I think a lot of them, my understanding, are getting so overwhelmed with all the things they have to do that there's almost a sort of paralysis that also sets in. They don't know what decisions to make, right? You have CISOs who have been inundated with money. That's chief information security officers. Perfect. Thank you. Yes, that's perfect. They are being inundated with money because their boards are finally figuring out, oh, cybersecurity really matters. And they're hearing about these fancy models from Anthropic and OpenAI and going, here's some money, go get that.

17:00And then you have security leaders who are now going, oh, my God, all the attention and money I've ever wanted is here. Great. What do I do with it? Like, how do I allocate all of this? Do I hire more? Do I get these models? Can the models even really help me? Do I need more specialized tools? And they have a backlog of things that they just were never able to really get to because they didn't have the money until now. And they have to go at a pace that is like so rapid. And maybe AI can help. Or maybe they just have to do the basics and do the grunt work really quickly. Okay. So those are private companies with boards, proactive.

17:40I think I just saw – this isn't coming out for a while. No. I just saw some data, you know, the CEOs are worried about cybersecurity more than anything else. It's crazy. What about the state, local and federal governments who have so much of our sensitive data, so much information online, so much tech capability that is already, there have been a few incidents already, right, where there's been hacks. I read about something over the weekend. Philadelphia had some fake criminal report called in by AI. There's been other things. What has already happened at the government level? And then we'll talk about response after that.

18:21Yeah, there are two things that come to mind. One, I will say you're totally right to ask about this. I think this is the area where I am the most concerned, right? As you were talking about entities that are dependent on state, local, maybe even federal offices to be able to fund them, right? So not only private companies are dealing with a board, red tape, et cetera, now you have voters who are figuring out who to elect or what tax measures need to go through to help fund, you know, whatever infrastructure developments are doing. And then you have water systems and power grids and transportation agencies all fighting for whatever money is coming through those increased taxes, right?

19:03And so, or through the budget or you're fighting with schools that also need funding and things like that. So there are a lot of concerns there. The Philadelphia thing, that was Anthropic's model during testing went a little bit rogue and called in a fake murder tip to the Philadelphia police. It is a bit ridiculous to me how that happened. Luckily, it landed in the spam inbox. Like the police didn't follow up on it. Like it was flagged. So it could have gone a lot worse, but it definitely is eye-opening. And it's among a wave of different things that AI models have been doing. They've been targeting government websites and kind of accessing data.

19:43They've been exploiting holes that, frankly, are pretty basic things to exploit. But then you think about the fact that AI agents were doing this and model testing. Like, it wasn't a hacker that was telling a model to go do this. They just were looking for the answers to a test and went, you know what? I think the census has this information for me and like broken that way. So there are a range of different things that state and local governments, even federal, I just brought up the census bureau that they have all been wrestling with for eons. And it is a lot harder to solve this stuff, especially when you're talking about critical infrastructure that relies on physical devices, right?

20:24So think about your water systems. You can't just go and hit patch or software update or whatever to get it to update. There is some equipment where you have to be physically on deck to install the patches in person. And that adds another barrier when you're talking about all the bugs that you're going to find or all the updates you have to make. Then you need someone who can be there to physically update it, to ensure that this water system can't be hacked or if there's a zero day in there, it can't be accessed and no one can tamper water. So it is a real concern. And then you pair that with nation states that have been targeting critical infrastructure for years.

21:07Iran has targeted our water systems. There are concerns around China kind of infiltrating our telco networks or pre-positioning inside of various critical infrastructure in case we ever want to interfere with an invasion of Taiwan, right? Then immediately they can start kind of messing with our infrastructure and then add AI to it. And it gets really scary really fast. and I can sit here and give a lot of scenarios, but it is top of mind for a lot of national security officials, AI labs, cybersecurity companies that I talk to. And I don't know that there's a great solution to it. It is only somewhat recently that I've started even hearing AI labs go, oh yeah, we're standing up programs to help state and local entities now.

21:55And you're like, well, this has been an issue for years, but cool, at least we're doing something now. So it's a real issue. I think about it a lot. Like if you had to pick, what are the one or two doomsday scenarios that like the people you're talking to, do they have like a top doomsday scenario that they worry about? It's a bit tricky because getting cybersecurity people to tell you their top doomsday scenario is really hard. These are people who are acutely aware of the power of their words and headlines, right? If they talk to me and then I go XYZ cybersecurity firm or this government official, this is what keeps them up at night, it grabs headlines in a way that maybe they don't want.

22:42I have heard privately of some of them being concerned about AI being used to target maybe one particular product that is used across one critical infrastructure industry, right? So think about a device that is maybe like really random that water systems all seem to really need. It's like super niche. I'm trying to think of a good one off the top of my head. But and then think about if there is a flaw in there, right, that you then can use to infect dozens of water systems. There was actually a series of water cybersecurity attacks that happened earlier this year in Minnesota and across the Midwest.

23:28And there's suspicion from the U.S. government that it was Iran that did it. They targeted like 30 water systems all at once. And luckily, no one was injured. There weren't water supplies that were deeply, deeply impacted. But in some regions, they had to do like boil water advisories because they were worried about the ways in which maybe the pressure in the water system was changed, right? And they just weren't sure. So you can kind of think about that and then go from there. And the second one that has come up probably as well is the concern around using AI for decision making, right? So there was an incident about a month ago, but definitely earlier this year at the Pentagon, where they had an AI model that hallucinated information.

24:16And we almost went to war with China over it. And it took like a human making a decision very last minute and going, wait a minute, guys, I really don't think this actually happened. I think the model hallucinated, but like we came scary close to making a choice at the Pentagon that like could have deeply altered like how we were or the geopolitical landscape that we're in right now. So it's kind of crazy. It's stuff like that. It's like the very minor things or things that you don't think about until they happen that I just don't think we're quite ready for or haven't fully wrapped our heads around.

24:53That seems like a new threat. Like you're saying AI sort of like puts regular efforts to hack on steroids and makes hackers more capable. But then there's another threat, which is AI makes it easier to stop relying on your own intelligence and like lend it to computer super intelligence, whatever people want to call it. And that's very scary and risky. Like you don't want to put decision making power into the AI's hands. Or maybe you do, or maybe that's something people are thinking about or worried about. I think people are thinking about it. I think that is a threat that I think deeply changes depending on your company and where you work, right?

25:42And how critical or sensitive your use cases are, right? I'm thinking of a variety off the top of my head. And perhaps national security issues like in military operations, high on that list, right, of areas where this is a huge concern. I'm thinking also, given the podcast I'm on, I'm not thinking of like, you know, if I worked at like the New York Stock Exchange, that would be a huge issue as well, right? There are places where it could turn into a really detrimental place or a detrimental consequence if you lean too much into trusting AI. And I think that's a big reason why you're seeing a lot of companies that are adopting AI agents in particular thinking a lot about where humans need to sit in overseeing what the agents do.

26:33And they're also thinking a lot about safeguards and transparency and things like that to ensure that it doesn't go off the rails. It's still similar, though, to – think of it like managing your employees, right? Like you kind of have to have visibility into what I can access on a computer. So it's kind of new in that, yes, we're thinking about trusting a computer to do a lot more work than usual. But securing it is kind of similar in that you're thinking about how you would monitor for, you know, a rogue employee trying to access paycheck information. Well, how do we put safeguards to ensure that, you know, there's a reason why Axios, I can't see, you know, everyone's expense information or data, right?

27:20So there are still solutions and it's still kind of analogous to similar security issues if that gives people hope. there is typically some traditional security measure that like will still apply in the AI world. It was just a matter of being faster, smarter, and perhaps even using AI to like fend off the AI that's coming for them. Okay. Last question. I'm going to put this one to Maria. You talk to people every day about all this stuff. How worried are you? Like DEF CON, if DEF CON 5 is like, we're at nuclear war, you know, are we at DEF CON 1? Everything's chill. Are we in the middle? How, how worried, how optimistic are you?

28:07Yeah, I'm interested to see how this changes in, you know, two days to a week. I think if you asked me last week, maybe, maybe I'd still be in the same place. I think I'm at a DEF CON three, in part because I still sound paranoid to your point, Emily, but I think in part, I would be more worried if the existing security controls that we had couldn't mitigate so, so much, right? And I'm still firmly in the middle because so much of this depends on organizations moving fast enough. And I don't know for sure that, you know, all the panic and fear that they feel is actually translating into actions and spending and strategizing currently.

28:56I think we're all understandably still like stuck in the lens of like fear paralysis and like, oh my God, the end of the world is coming. Should we just like get in our vans and go West, like, I don't know, like, what do we do kind of vibe. So I think DEF CON 3. But let's see what happens in a week. All right. Sam, this is awesome. This has been great. Thank you so much. Thank you.

29:27So I want to pivot now to talking about the White House response to AI. So we're going to take a quick break. And when we come back, I'll be joined by Sam's colleague, Maria Curie.

29:43Slate Money is brought to you by Charles Schwab. Decisions made in Washington can affect your portfolio every day. But what policy changes should investors be watching? Listen to Washington Wise, an original podcast for investors from Charles Schwab, to hear the stories making news in Washington right now. Host Mike Townsend, Charles Schwab's Managing Director for Legislative and Regulatory Affairs, takes a nonpartisan look at the stories that matter most to investors, including policy initiatives for retirement savings, taxes and trade, inflation concerns, the Federal Reserve, and how regulatory developments can affect companies, sectors, and even the entire market.

30:24Mike and his guests offer their perspective on how policy changes could affect what you do with your portfolio. Download the latest episode and follow at schwab.com slash Washington Wise or wherever you listen. Slate Money is sponsored this week by Your Next Dollar, a podcast from Nerd Wallet Wealth Partners. Your Next Dollar is built specifically for high earners like you. It covers the financial complexity that comes with a higher income. If you have equity compensation, RSUs, tax planning, housing decisions, figuring out when it's actually okay to enjoy your money, it turns out that the more you earn, the more complicated the questions get.

31:04And yet, somehow, there's even less content built for you until now. Your Next Dollar covers both sides of the equation, building wealth and spending on what actually matters to you. Because at a certain income level, the question isn't, should I save more? It's, what should I do with everything I have? Financial advisors from Nerd Wallet Wealth Partners join the show regularly to discuss the questions they hear from high earners like you, the nuanced real-world stuff that generic financial content doesn't cover. If you want help getting more out of every dollar you earn, follow Your Next Dollar wherever you get your podcasts.

31:43Sponsored by Your Next Dollar, a podcast from NerdWallet Wealth Partners and SEC-registered investment advisor. This podcast is for informational purposes only and is not investment, tax, or legal advice.

31:58Welcome back again to Money Talks. And now we have the amazing Maria Curie, who is here to talk all things White House and tech policy with regards to AI security. Maria, welcome. Hi, thanks for having me. I'm really excited to talk to you because I kind of touched on this with Sam, but basically, you have the whole private sector kind of starting to freak out about security threats posed by AI. And then you have the public sector, you have governments kind of now needing to deal with this as well. And I feel like you are just right on the forefront of this in the policy trenches. I mean, President Trump sort of came into office wanting very little regulation, get the government out of all of this stuff, get the government out of AI's way.

32:50And they still talk like that. But lately, it seems like the administration's sort of getting sucked into doing something about AI. Like what has sort of changed in this dynamic, the push and pull? It's fascinating. And it's what continues to amaze me in all of this reporting is how much of these decisions that are consequential for huge swaths of the population and for even humanity, you, right, that's how we talk about these things in these terms, are being decided and driven by like personality and by people liking each other and by attitudes toward the industry. And so, you know, just today I was talking to some industry sources that were saying, you know, under the Biden administration, it was harder.

33:40It was harder for the AI industry. And we felt like the Biden White House really viewed us with so much distrust and contempt. And they really didn't want to talk about AI. They didn't see us as champions. And they didn't want to kind of support us in this shared goal of safety. And so the communication between the Biden administration and industry was so fraught. And now the way that industry views the Trump administration, it very much is a hands-off regulatory approach. The eye industry, these companies, they are our champions. You see these scrums. I was there in that scrum with President Trump and all of like the biggest tech CEOs.

34:25I think that was two weeks ago at this point. And just like the buddy-buddiness of it all between all of these people. And so in that sense, the Trump administration is a hands-off regulatory supportive approach. But this is also the administration that forced the shutting down of anthropics models from the internet where U.S. citizens could no longer access this technology, foreigners within the U.S. couldn't access this technology, internationally allies couldn't access it, and of course adversaries can't access it. What you would think is hammer down regulation, government intervention, if the Commerce Department sending a letter shuts down one of the world's leading AI companies, if that's not government intervention and regulation, but it's not perceived that way by the industry because of the way these things are talked about and the general attitude around it.

35:28And so what we saw, the latest thing that we saw was when President Trump's new AI task force being led by AI czar Jay Clayton. He has other jobs, right, Maria? Like Jay Clayton is also... The director of national intelligence. Yes. And so they are now mandating, it's no longer voluntary, they are mandating that when there is an AI incident, these companies share documents and share information with the government. That's new. That seems... When you're talking about the White House shutting down anthropics model, It seemed like very one-off and spiteful because it seemed like the president maybe had an axe to grind perhaps with the company CEO.

36:17It seemed a little spiteful and one-off. And I think around regulation, people want stable rules that they can kind of follow if you're going to do something. So is this latest move, is this a small step towards a stable rule regime something? Or is it just like a little, is there any enforcement? No, there's no enforcement and there's no implementation details, which has been the case, I think, throughout the entire second term and President Trump's approach to AI regulation and his administration's approach. I think what we're seeing with Anthropic, it's both things are true. There is a spitefulness to it.

36:56There is just the fact that the administration doesn't love the CEO and it's a midterm year and political foils are effective campaign strategies. and so that is happening but it is also happening that what happened to uh to claude and to anthropics models very much had implications for the rest of the industry and so it wasn't only going to be anthropic whose models were ever going to be shut down if the entire industry didn't get in line open ai's models soon after they did on their own decide to delay the release of their models and now they all have to adhere to a pre-deployment testing regime with the government because of the aftermath of what happened with Anthropic.

37:43So Anthropic continues to be the easy target, but whatever comes out of these fights is going to be applicable to the entire industry. We talked to Sam about just like how companies are sort of looking at their systems and like looking at them with a fresh eye because AI makes hacking so much easier now. And is the federal government doing that also? Like, you know, in addition to thinking about regulating AI, just looking inward at all the systems and data and all its vulnerabilities. Is the White House doing anything with that? Right, like the whole government, like, the whole like government modernization of it all, which we've been talking about since like Obamacare and like healthcare.gov.

38:29I mean, yeah, again, like, look, I don't want to be negative, but we're seeing a lot of rebranding ai to super intelligence like hey like messaging is important for the psyche and for elections and it could be impactful for the electorate tbd but like government modernization i think falls under this bucket too where we where we have a website now we have a website from the white house that is going to make government services like filing your taxes more efficient in the next year. That's not implemented yet. They rolled out the platform, but you can't right now actually go onto this website and file your taxes more easily or get your benefits from the VA more easily.

39:16That's something that the government has been trying to do even before President Trump came in. And then we also have what we have reported on together, Emily, and that you've reported on yourself, which is like Doge's impact on the government's workforce and the lack of technical expertise that we are now viewing in the government. How bad is it? Pretty bad. And like, you know, it depends on who you ask if it's bad, because the Trump administration's approach is intentional. They want to outsource the expertise to the industry. They don't see any issues or any conflict of interest problems with taking the industry's lead on this.

39:56and you do see a lot of like the government's top experts being poached by the industry. At the same time, you have the Office of Personnel Management, Scott Cooper, trying to hire those 1 ,000 people, those 1 ,000 tech experts by the end of the year. Last time I checked in with him, he has about 330-ish. If he gets to 500 by the end of the year. That's a success in his book. He doesn't, he's not pressed. He doesn't feel like I was at a lunch as well, where I was talking to people at the office of patent and trademarks and another person who works in an intelligence community agency. And they were saying, yeah, like we are modernizing, we're using AI and we have like the old guard employees where that are resistant and they don't want to do it.

40:47And like the compliance, but of course them, these staffers that were at this lunch, they were going to be excited about it all. And so the internal dynamics are definitely happening at these agencies for sure. But it's a weird thing to be having these conversations as the cyber attacks become more real, because that's obviously some of the most sensitive data that we have is at these government agencies. So I don't know that we want to modernize too quickly without appropriate safeguards. Right. And then just talking to you about what lawmakers are doing and what regulators are doing. Congress, anything there?

41:24Are we seeing anything happening? So they I saw I think it was maybe you or maybe our colleague had a story where he you guys asked lawmakers like, do you use AI? And they're like, No, we'll get to it. Oh, yeah, that was awesome. That was awesome. That was stuff. And Hans. Yes. Yes. Steph Kite and Hans Nichols. Yeah. That's right. That story was awesome because it shows and they know this. It shows that like the lack of familiarity with the technology that these lawmakers should have. It is also true that you are not going to solve AI policymaking's trickiest questions by using ChatGPT. That requires talking to experts outside advocacy organizations and getting insights from people outside of industry that have their own motives and incentives, especially ahead of IPOs, right?

42:20So Congress, what is Congress doing? Congress is campaigning. Congress is trying, Democrats are trying to get a majority in the House. And let's say they do get a majority in the House. If they can agree amongst themselves on the specific policy solutions to AI safety, they still need to get the president to sign it. And why would Democrats want that type of win to be under a Republican president? Why not just wait until 2028? Maybe you have a trifecta, and maybe the issue is just bipartisan enough to where you can get enough Republicans on board and you don't need a trifecta. The other thing that our reporting has shown too, in terms of the way that industry is thinking about this, is that they don't expect Congress to do anything about it until there is a catastrophic event.

43:08Yes, I wanted to write about this. This is sort of like the basic theory of how policy gets made, especially around new technology. Nothing really happens. No one knows exactly what to do. And the catastrophe is the clarifying moment where you're like, oh, now I know what to do. We have to do whatever will prevent that from happening again. Yes, yes. And it's like the forcing mechanism for Congress to finally do something, and I talked to Hill staffers who were like, yeah, during the pandemic, we were able to come together and get our act together. After the 2008 financial crisis, we were able to get it together.

43:44But here's why I would take that approach with a massive grain of salt, especially because it's coming from industry players. States are already regulating, and they didn't need to have some sort of catastrophic event. They didn't need water supplies or the power or the internet to go down before they put basic transparency measures in place. And I do think that the lobbying we're seeing at the state level is powerful. It's slowing things down. And I don't think that we have to just take as a given that the only way Congress or that the only way we will have regulation in the United States is waiting on Congress and waiting on a catastrophic event.

44:29State houses are acting now and also lawsuits are happening now. And all of that is already forcing change. What are the states doing? Transparency bills. And they keep building on each other. So you have essentially requirements, mandates, not voluntary, where whenever an incident happens, these AI companies, they have to share what went down with third party auditors, independent verifiers, relevant government agencies. because it is in that like reverse engineering of what happened and in those independent eyes that you can better understand and then prevent the next thing from happening because a lot of the times these systems and sam knows about this way more than i do but these systems behave in such a way where like human beings can't actually understand what happened and you need a lot of different eyes and a lot of different experts to come in you need also international pockets of expertise with AI safety institutes in the UK, in Japan, like all over, there are a lot of experts that are working on this, but it's a matter of having that transparency mandated by the government because the companies are not going to have that incentive to, for lack of a better word, air their dirty laundry to the world.

45:46Like that is why you hear like Amadei saying you shouldn't just rely on me and my goodwill because like I am a company CEO and I do have to succumb to market pressures ahead of an IPO. So what have the states done? Transparency measures. And it's interesting, they keep getting stronger than the next. So the next, I think, California passed something, and then Illinois passed something even stronger. And now Massachusetts is on the verge of passing something even stronger. And the bar just keeps getting higher. And the thing for the federal government to preempt or to get in the way of is just becoming even more and more cumbersome.

46:22So is there lobbying going on to get the federal government to preempt state law? That's when basically the federal government says it doesn't matter what the states require, like our law supersedes them. Yes, the lobbying is 100 % happening toward a federal standard that preempts state laws, all in the name of giving industry something better than a patchwork of cumbersome state regulations that's difficult for compliance and not in the interest of innovation and winning this technology in a fierce race against China. It's the same playbook we saw with privacy and why today we're one of the few developed nations on earth that doesn't have a federal privacy standard.

47:08Great. So we're messing up AI policy just like we messed up social media policy. I guess that that would be the conventional wisdom. Is that accurate? I think that is what people would say. Well, I should ask you the same thing I asked Sam, since you're immersed in all of this, how like what's keeping you up at night? How worried are you in terms of sort of like AI security risks? Are we at DEF CON 5, which is like nuclear war? Are we at 1, which is like everything's chill okay i have a different take on this i actually i'm not worried or scared i'm really not love it i am actually i am energized by the way in which technology is mobilizing the electorate one way or the other like i'm not taking a position on regulate it don't regulate it i just see the ways in which communities are engaging around like the data center issue and the way states are moving forward with regulation the way candidates are increasingly talking about this as like a central issue of their campaigns tying it to these bigger issues around affordability around the economy around corruption uh the influence of money in politics and i think that we are seeing this shake up conventional politics and conventional norms that could actually end up resulting in a great outcome for society.

48:37I think it's just like an open question that we're going to see in November. We're going to see ahead of 2020. I think November will be the first test. And yeah, I think about that a lot more than I think about the internet being shut down. That's great. Maybe that'd be a good thing. Maybe it wouldn't be a bad thing if the internet were shut down so we can get offline for two seconds and touch grass. Yeah, we all need to touch grass, but I don't want our water systems getting messed with. Yeah, we do need water to survive. That's true. Maria, thank you so much for doing this. I really appreciate it.

49:15Thank you. This was fun. That's our show for this week. Thanks to Jessamyn Molly and Shana Roth for producing. Ben Richman is Senior Director of Podcast Operations. Mia LaBelle is Executive Producer of Podcasts. I'll be back in your feed on Saturday along with Felix and Elizabeth for a regular episode of Slate Money. And until then, thanks for listening.

49:59to the work behind the work. We're talking about matching receipts, chasing card charges, managing travel, nearly an entire workday spent doing administrative chores nobody was actually hired to do. It's a massive drain on productivity, which is why Perk exists. Perk is the intelligent platform for travel and spend. Instead of suffering through five different tabs in a broken spreadsheet, Perk combines business travel booking, expense management, and corporate card control into one single system. Your finance team gets real-time visibility and everyone else gets their sanity back. Save yourself the headache.

50:32Sign up at perk.com slash slate money to enjoy no platform fees for the first three months. That is perk.com slash slate money. New U.S.-based corporate customers only terms apply.

50:55shrink facility. You're balancing inventory and budget while sourcing critical supplies across the operation. Too much inventory? Costs climb. Too little? Production slows. Grainger's$2 billion in on-hand inventory helps you get what you need when you need it, so you can help keep budgets balanced and lines moving. Call 1-800-GRAINGER, click grainger.com, or just stop by. Grainger, for the ones who get it done.