The AI election

Megan Rapino here. This week on A Touchmore we're bringing you our live show in Phoenix with WNBA 4-time champion Chelsea Gray and the Naismith Coach of the Year, Shay Rau. Together we talk about the NCAA semi-finals, the crazy activity in the transfer portal, and of course the final matchup for the NCAA championship. Check out the latest episode of A Touchmore, wherever you get your podcasts and on YouTube. What should we make of the Iran War ceasefire announcement and where do things go from here? If anything has surprised me over the last 24 hours, it's that Iran agreed to a ceasefire, and particularly that Iran agreed to a ceasefire after that outrageous message that President Trump put out. I'm Jake Sullivan and I'm John Feiner and we're the hosts of the Long Game, a weekly national security podcast. This week we break down the latest news on Iran and share our net assessment of where things stand for the US. The episodes out now, search for and follow the Long Game, wherever you get your podcasts. This is The Weeds, I'm John Quinhill. A couple weeks ago we got a question from a familiar voice, at least to me. If you listen to the show regularly, it's a name you probably recognize. Hey, this is A.M. Hall, editorial director of The Weeds, and just like you, I have questions, and there's one question that's been nagging at me lately, and I'm hoping we can answer it today. For me, and for all of you, here it is. On a scale from 1 to 10, how worried should I be about AI and the election? A is the only one with this question. It's something I've had my eye on too, especially this year. 2024 isn't just an election year here in the US. This year, we'll see over 50 other countries around the world hold elections, impacting half of the world's population. It's going to be a real test for democracy, and we're already seeing the ways that ever changing technology is playing a role. From chat GPT to image generation, audio deepfakes, and more. To find out what's going on with AI campaigns and what this means for elections, I made a call. I'm Tiffany Shoe, and I cover misinformation and disinformation for The New York Times. So I put the question to Tiffany. On a scale from 1 to 10, how worried should we be about AI in this election? I would say be nervous. A lot of the experts that I talked to were quite concerned, and they spend their careers thinking about this sort of thing. So artificial intelligence, AI, broadly speaking, how can AI technology be used? In an election context, in a lot of different ways. It's really a risk-reward situation with AI, especially this year. A lot of people are saying that AI could be helpful in reducing campaign costs, because it can essentially create instant responses to the bait questions or attack ads. It can be used to quickly analyze data. Apparently, some politicians are already using it for translation purposes. So for example, I'm a politician. I put myself in a campaign ad as a video, and then I can distribute versions of that video where I'm speaking languages that I don't speak in real life. And so the idea is you can get closer to underserved communities or communities you wouldn't otherwise be able to reach. So that's the pro. The con is that because AI can be used to create really any kind of content, we're talking text through chatbots, we're talking image generator, so anything visual that also creates deep-fake videos and audio, it can be used to deceptive ends. So the concern there is I could see a video of myself being circulated that was never actually filmed because it was created by a machine. There could be emails sent out at scale, quoting me saying things that I never said. So as a vehicle for spreading lies, it could be pretty effective. That's really scary, especially, you know, our voices are out there. I don't know, that's something I think about a lot, even beyond the political implications. Right. I have had so many researchers explain to me that you can make an audio spoof in minutes. There was a scam going around over the holidays involving Taylor Swift, where her voice was spoofed to appear as if she was selling La Crusade cookware. She obviously wasn't, but it sounded like her voice. And if you were scrolling through a social media platform quickly, you would probably be done the wiser. If you apply that to an election situation, that's pretty dangerous for democracy. Yeah, I think around that same time I saw a video floating around of someone spoofing Cardi B's voice. Oh, it's a girl cardie saying like, oh, the government is giving these types of refunds. And I'm going to show you how to put $16,000 in your pocket right now for free. It very obviously to me did not seem to be Cardi B, but someone somewhere may fall for that. Yeah, I mean, there's an example of this already potentially having an effect in an election. Last year there was an election in Slovakia where audio that most researchers now believe was likely a deep bake spread that seemed to show the pro-Western party candidate claiming vote buying that might have played a role in that party losing to a pro-Cremlin party. That's Slovakia example is being cited pretty often now as potential evidence that audio was something to be concerned about. But there are many, many other examples already of deep bake, of text, of audio being used to likely malicious ends in the elections that are happening around the world. You mentioned before some of the ways campaigns may want to use this technology, whether it's reaching people who speak a different language, mass emails, are campaigns embracing the technology or are they kind of issuing it? There are mixed feelings about it. There are definitely campaigns that are doing AI-related work and have been doing it for years. For example, chatbots are a pretty simple way to draft emails, like fundraising emails. It's not that hard to use a chatbot to come up with a couple of different versions and then maybe have a human edit it. The DeSantis campaign already produced AI-generated imagery of Trump embracing Fauci with no disclaimer. The RNC had an ad that showed parts of America and scenes of chaos. There was a Toronto mayoral candidate that had a campaign featuring fake dystopian images of homeless encampments. My favorite was an image that showed a woman with two arms crossed and then a third arm touching her chin. It was pretty obvious and sloppy, but it was still a use of AI in an election. In New Zealand, there was a party that posted on Instagram this rendering of robbers in a jewelry shop and it looked pretty realistic. Dr. images and political propaganda have been around for ages, but why is this moment different or more dangerous? Is it? Oh, absolutely. Because now you can lie easily and at scale. In the past, if you wanted to, for example, generate a blog post that spread false rumors about a candidate, you would probably have to write it yourself and then post it yourself. You don't have to do that anymore. With chatbots, you can do multiple versions very quickly off of a service that is probably free or accessible at a very low cost. It doesn't take any great technical know-how to use this sort of thing and then you can post it wherever you want. In the same thing goes for audio, to a lesser extent, the same thing goes for image generators. A lot of the AI companies that are producing these sorts of tools are trying to create guardrails. They're making it so that certain prompts don't produce imagery or text or audio that could have a negative effect. But it's also pretty easy to get around a lot of those guardrails. Yeah, I want to go back to this idea of AI being more dangerous now because of the sophistication. How is that changing things? How sophisticated have gotten? What are we saying? A few months ago, a colleague and I did a story where we asked, can we no longer believe anything that we see? We looked at image generators and their ability to produce realistic images. At that point, there were still certain tells. For example, a lot of the generators tended to produce people with multiple fingers, multiple extra fingers, or fingers that were weirdly stubby or floppy. It was like everything everywhere all at once, but in an AI image, a lot of the generators also had issues with producing text within an image. You looked at renderings of people in the background of certain images, a lot of them would start to meld together and look blurry. That is increasingly no longer. the case, a lot of the chatbots and image generators and voice cloning tech has gone through upgrades in the past year and they're much much much better. The added concern is that the detection technology is not catching up and it really never will because detection technology by its nature is designed to react to existing technology. So you're always going to be in this cat and mouse cycle where the AI tech is more advanced than the technology that exists to try to detect its use. And so the problem is now that you have AI technology that is capable of producing content that is extremely and increasingly realistic and it's not easy to know that. There are lots of elections around the world this year and you know we've already seen some. How have we seen AI used abroad so far? I wish we had hours and hours and hours to discuss the examples because there are many. There was an election in Argentina last year where AI was used pretty prominently. Candidates used it to portray themselves as heroic figures like soldiers, a ghost buster, Indiana Jones showing them on like the cover of the New Yorker and then it was also used to show their opponents as like the villain in clockwork orange. India is preparing to vote. Their elections are in the spring. This is the world's biggest democracy. Their information technology minister Ashwini Vanshna called deepfakes a quote threat to democracy which is a concern that shared by their prime minister on Narendra Modi. There have been information, influence operations that target elections. There's a campaign that's known as spam a flage. It's actually been around for a few years but researchers recently have said that it seems to be pivoting to target the US election in November. In part by using AI images that show like President Biden and former President Trump like in battle stance with spears that are on fire. It shows images of like cracks in the white house and the statue of liberty. On Twitch there is a 24/7 deep fake debate between Biden and Trump. Oh my gosh. I saw that. It's wild. If you're just like if you're glancing like if it was on your TV and your living room and you're just walking by you would be like oh there's a debate today. Totally. If you if you told me that is real I would not say oh you're crazy because it looks like it could be real even though they're saying completely ludicrous things that I think not even they would say in real life. You think I'm scared? I eat mega clowns like him for breakfast without even breaking a sweat. That clip you just heard? It's a bunch of malarkey. Up next what everyone from states to tech companies are doing to curb deep fakes. Support for this show comes from factor. It can be hard to keep your healthy eating habits when the grill start coming out. You can almost smell the hot dogs from here. But if you want to keep eating healthy throughout the year then maybe you should look at factor. Factor prepared meals are designed by dietitians and crafted by chefs. Ready in two minutes no planning no cooking. With 100 rotating weekly meals to keep things fresh and delicious factor meals that fit your goals and schedule. Our collie Claire White got a box from factor and here's what she had to say. What I loved about factor was that it was ready to eat right out of the box. All of the different variety was delicious. They have so many options for your different dietary preferences and they make it really easy to find the nutritional information of each meal. Plus factor meals are always fresh. Never frozen with no prep, no stress so you could actually stick to your goals. Head to factor meals.com/explainit, 50 off, and use the code "explainit 50 off" to get 50% off and free breakfast for a year. Offer only valid for new factor customers with code and qualifying auto renewing subscription purchase. Make healthier eating easy with factor. Hi, I'm Bernabé Brown and I'm Adam Grant and we're here to invite you to the Curiosity Shop. A podcast that's a place for listening, wondering, thinking, feeling, and questioning. It's going to be fun. We rarely agree. But we almost never disagree and we're always learning. That's true. You can subscribe to the Curiosity Shop on YouTube or follow in your favorite podcast app to automatically receive new episodes every Thursday. How is Trump's psychology having an impact on the great power conflict? There were folks who for years could never imagine the US carrying out limited strikes on Iran. If you go back to the 12-day war, he dropped those bunker busters, right? And you had presidents through multiple administrations who never would have gone to full scale or with Iran, and here they are. I'm Pete Barara and this week CNN's Chief National Security Analyst, Jim Sudo, joins me to discuss the Iran War, our fraying alliances, and the rise of Russia and China. The episode is out now. Search and follow Stay tuned with Preet, wherever you get your podcasts. Okay, what is the regulatory landscape currently as it relates to AI and elections? I want to start at the state level. What are we seeing there? Yeah, so this is something that has been pretty surprising to me because generally you don't see legislators at any level moving with any amount of speed. It kind of seems like this year a lot of them realize, oh, we have kind of a major election coming. We should probably figure out this AI business. So as of January, 32 states have bills to regulate the fake elections. Five states have enacted laws that's California, Texas, Washington, Minnesota, and Michigan. They all have varying penalties. Some are more stringent than others. Lots of questions about whether or not any of those regulations can effectively be enforced because so much AI content generation happens anonymously. But just the fact that there's so much movement so suddenly is telling as to the level of concern about this tech. What about the federal government? What does regulation look like there? As far as I know, regulation doesn't exist at the moment. There is a bipartisan effort in the Senate from the Senators Kobushar and Holly to try to draft regulation related to the use of AI and federal elections. There's another effort in the House from Yvette Clark. They're still discussing it. I think a common worry that's voiced by a lot of experts in this field is that AI is an extremely complicated technology and a lot of legislators don't understand it. They don't understand how quickly it's advancing and how much of a heave would be required for legislation to keep up with those advancements. There are also some other federal level efforts. The Federal Communications Commission this month outlawed unwanted robocalls that were created by AI and that was a response to robocalls that mimicked President Biden's voice during the New Hampshire primary in January. There's been some criticism of the Federal Election Commission for dragging its feet because they haven't produced any rules, specifically about AI campaign ads. They're saying that they might make a decision by summer but then honestly it's probably too late. The election was like that like we're more than halfway there. Right. And I heard that robocall. I was surprised by how realistic it sounded because a lot of I mean in the past a lot of them sounded fake but that one I'm like okay I could see how someone should get fooled by that one. Yeah the audio technology is frankly what concerns me most. With AI imagery there are at least visual cues that you can sometimes pick up, you know like the hands like the text. With audio it's really tough. Sometimes the audio comes out sounding like a little robotic. The pauses might be a little weird but if you're listening quickly who's to say that something is real or not real. Earlier you brought up lawmakers and the fact that you know this is complicated technology and lawmakers you know they're notoriously not the most tech savvy. I think of Senator Blumenthal's like Finsta Blunder. Will you commit to ending Finsta? And that's a major reason why when it comes to tech companies you know the government can be kind of slow to act but it seems like especially on the state level people are moving quickly. What does that signal and you know is there concern at all that they're moving on things that they don't quite understand. That absolutely is a concern. I think there are different camps in regard to the speed of legislative efforts around this. There's one camp that says let's just get legislation in the books. Let's just try to have guardrails in place. We can iron out the specifics later but we've got to start policing this sort of thing. There's the other camp of course it says hold your horses. We don't really understand how this technology works. We don't really know how much of an effect is going to have on elections. Let's to study it and produce a law that is coherent and comprehensive. That said, there is some relief among the experts that I talked to that the tech companies themselves seem to be increasingly invested in this sort of thing. A number of companies have in the past few weeks released rules regarding AI use in elections using their tools or their platforms at the Munich Security Conference, 20 companies, and that includes companies like Amazon, Microsoft, TikTok, and even X, which is not particularly known at the moment for content moderation. Yeah, it's a wild west over there. Yeah, but they all sign this pledge to keep AI from disrupting voting, which was, it's probably a symbolic move, like who's to say what specifics are actually going to be able to enforce. But the fact that 20 companies got together and were like, we should address this is pretty telling. And then there have been other guidelines and suggestions released by like OpenAI, Google, Meta, Anthropic. What's funny about OpenAI is, you know, they're announcing rules to prevent abuse of their tools during elections. And then they're releasing Sora, which is this like phenomenally effective AI-generated video tool. It is like stunningly realistic what Sora is able to do. Recently, some of the sample videos came out and you know, they probably aren't going to show the crappy ones. But it was shocking, like it looks like actual film. It's funny though that you say they're not going to show the crappy ones because a number of the companies when they were releasing their chatbots or introducing their chatbots showed their chatbots hallucinating and sharing fake information were inaccurate information. So a lot of these releases aren't as a pristine as a tech company might hope. Yeah, are the tech companies sort of the ones in the driver's seat when it comes to this at the moment, especially because you know, they're developing the technology, they're in control of it. I'm just curious sort of like, what's the calculus there and what thereafter? I mean, there's a lot of hot potato when it comes to who's responsible for air regulation. You know, you've got some side saying, oh, it should be the government, it should be regulators, others are saying, no, it should be the company is like anthropic and open AI that actually produce these tools. And then others are like, no, no, it should be the social platform. It should be meta. It should be tick-tock because they're the ones that are helping to spread most of the content that gets produced. And then there's also a camp that is like, consumers of content should improve their media literacy so that they can catch it. So who's to say who is ultimately responsible? It's probably a mix of all of them. But for the tech companies, I mean, AI is a lucrative business. This is an industry that is having just unfathomable loads of cash being pumped into it by investors. So they're all and have been in a race to produce the best and most effective technology out there before consumers start to winnow down to like an obvious winner. They all want to win. And so some of them are starting to tackle this question of potential harms. But profit continues to be a major motive here, if not the primary motive. The cynic and me thinks, you know, okay, they're trying to keep the backlash that maybe happened after the 2016 election and misinformation from happening. It's like, how can we get ahead of this so we don't have to deal with that again? There's probably a component of that. I mean, a lot has been written in the past few months about some of the major social platforms backpedaling on their election integrity efforts. I mean, X laid off a huge portion, if not all, of their election integrity team. Trust in safety teams and meta have taken a hit, same at YouTube. There has been a distinct pullback in how a lot of these platforms are prioritizing election moderation. This is maybe a way to balance that out with these AI-related efforts. But it's hard to say how effective a lot of this is going to be in the heat of a campaign. I know it's different for the people who are developing the technology. But even for those who are developing it, there's still so much unknown. Is it sort of the same cat and mouse game that we're seeing with detection software? You know, we don't know what an issue is until it presents itself. Yeah, I mean, the AI boom really didn't start until last year. We're talking about an industry that has seen incredible leaps and bounds in just a couple of months. So the fact that detection capabilities for the most part are terrible is not super surprising. My colleague and I wrote a story in June about detection tech. We tested some of these services and the results were astounding, frankly. I mean, there were images created by AI that showed Elon Musk making out with a robot where a bunch of the detection services were like, yep, real photo. There was an image of a giant Neanderthal that was standing next to like normal sized people. We looked at five different detection services. All five of them said that that Neanderthal image was a real photo. There was an image that purported to show a bombing at the Pentagon that was quickly debunked. I mean, the building was upon closer inspection, not the Pentagon. The smoke looked kind of funky. I mean, it was pretty clearly an AI generated image. Four of the five detection services were like, yep, real photo. So I mean, when you have detection services that can't tell that an image of Harry Potter actors in pastel, West Anderson style is not a real photo, then there's reason to be worried. And those detection services have probably improved since we wrote that story, but so have the generators that they're trying to detect. Up next, we'll see if I can spot the fake. When you think of a phone, not a smartphone, but like a telephone, you almost certainly think of one particular device. This boxy thing that's out on your desk, it has a bunch of buttons on it, it has a handset that you pick up, it has a braided cable, you're thinking of a phone called the Western Electric 500. And for decades, it was absolutely everywhere in American homes. It was essentially illegal not to have it in your house if you had a phone. This week on version history, our chat show about the best and worst and most interesting products in tech history, we're telling the story of the Western Electric 500, which is actually the story about the AT&T monopoly and how it fell apart. All that on version history on YouTube and wherever you could broadcasts. And we're back. It's the weeds. I'm John Guinhill and I'm here with Tiffany Shoe from The New York Times. Okay, Tiffany, before the break, we talked about fake image detection and I want to try something. So we're going to do a little test. I've given you the link and we are going to try to spot the fake AI photos. Just to describe what I'm seeing, it looks like a man maybe in his late white man in his late 30s, brown hair, a little bit of stubble. And now I have to try to see if he's real or not. I'm going to say that this is a real person. And I was wrong, it was generated by AI. Okay, we can try one more. Okay, this is a younger guy maybe like 20s. He could even be as young as 19. Dark hair, brown eyes. His facial hair looks suspicious to me. So I'm going to say he is AI. And I got that wrong. He is a real person. Sir, I'm sorry I said that your facial hair was suspicious. So that is concerning. I consider myself someone who is extremely media literate. But some of this is still really convincing. Like these photos, like that Biden, Robo call a new Hampshire. Have you been fooled before? How do you navigate this? Will I admit that I've been fooled on this podcast? I will say I have not been fooled yet in print and anything that I've published. I mean, I have seen many, many images that at first glance, I was like, yes, that could be a photo. I mean, frankly, I took this quiz and it was not pretty. It was not pretty. And every time I've done a story about how terrible detection technology is, I've had to look at images like this. And I'm like, you know what? You're also not great at detection. It's hard. It's really realistic. Why aren't companies, regulators, anybody? Why not institute a band of some kind on this technology until, you know, the election is over? I mean, they can certainly try. There have been efforts by some of the tech companies to be like, you know, you have to disclose this sort of thing or, you know, you can't use our tech technology or our tools to campaign. But how do you enforce something like that? A lot of the people who are playing around with this tech are anonymous. There are Reddit groups, Facebook groups where people are just sharing reams of AI generated content. And you don't know any of their names. They're all hiding behind anonymous account handles. So it's one thing to say you are not allowed to do this. It's another thing to say, we can stop you and hear how. Because I don't think anyone really knows at the moment how to enforce a prohibition on this sort of content. I am really curious. I'm thinking of not just media literacy, but AI literacy 101. What should we ask ourselves or be on the lookout for? How can we avoid being fooled by some of this really advanced technology? So I like to use this example of a panel that I watched. It was a virtual panel that I watched late last year. And it featured a guy named Larry Norden, who is from the Brennan Center for Justice. And he's an elections expert. And he was on this panel talking about some of the AI related concerns when it comes to elections. And when he finished, the organizers were like, Larry, we've got a surprise for you. And then they pulled up a deep fake of Larry. Using his voice to make some of the same points that he had just made. And the takeaway from that, I think, is A, that the technology is really good now. You can expect mimicry that will persuade a lot of people. That's real. But B, also that the solution is right there. And the solution in that case was Larry, right? All you would have to do is go to Larry and say, Larry, I saw this footage of you, is it real? And of course, caveat, Larry could lie. This is actually something called the liar's dividend. It's a phenomenon that a lot of people are worried about, where politicians could use the presence of AI to say, oh, that audio or that video you saw of me cheating on my wife while I was trying to rig the vote. Yeah, that's deep fake, even if it was actual footage. I mean, that's a concern. But what I've been telling people is it's probably a good idea to try to find authoritative sources and start from there. If you can go to state elections websites, if you can go to no nonprofits that deal with election integrity, if you can go to a candidate's website and look at their platform there, then it's a good jumping off point to generate your own analysis, to form your own opinions, and to make any decision that a voter would. Because then you know that you're starting from a point that is pretty provably authentic. Is it possible at some point for bad actors to spoof some of those websites and some of those authoritative sources? Yeah, there's always going to be a work around by people with malicious intent, but it's a good place to start. How hopeful are you that people are going to start kind of doing that due diligence when they come in contact with a piece of media in the wild? We don't have a great record when it comes to media literacy. And do you think that AI is going to be the thing to shift that at all? There has been research that offers a little bit of hope, that especially after 2016, that a fair number of people became wiser to the fact that there is information and content out there that probably shouldn't be trusted. Do I think people are more aware that there's misinformation disinformation out there? Yeah, I do. Do I have confidence that they care? Not necessarily. I mean, it's tricky dealing with the subject because while it has major ramifications for democracy, for how the world is run, for how we perceive truth and reality, it's a lot for people to process. If you're constantly telling someone, don't trust this information, be skeptical, you end up with something that a researcher wants described to me as information nihilism, where people don't trust anything, where they throw their hands and say, if I don't know what reality is, I'm not going to care to find out, which is dangerous in and of itself. So the balance is hard to strike in this case, but I think people are wiser to the fact that this is a problem. Actually, there is a good example in the case of Taiwan. Taiwan had their election last month, and in the lead up to it, there was a lot of chatter about how disinformation from China might sway the election, how it could negatively influence how voters perceive certain issues. But I think the conclusion reached by a lot of people who study Taiwanese politics is that actually in the past few years, Taiwanese civilians have become pretty literate in media. They make use of the fact-checking community that is there. They tend to take a beat before trusting some of the information that circulates in their media ecosystem. Unlike a lot of places, Taiwan actually has a pretty robust and pretty diverse media landscape, which is always helpful. And it was kind of a promising sign for a lot of people that in this particular election, which I should caveat by saying, wasn't especially close, people were nonetheless heartened to see evidence that disinformation didn't seem to be as effective as had been feared. What are the implications for all of this beyond elections and politics? What is the impact that AI could have on, you know, just that everyday person listening to this conversation? I mean, AI is used in everything already, right? It powers a lot of the recommendation algorithms that we use when we're searching for things, when we're shopping or whatever. It powers a lot of the map tools that we use. All of these random technological implications are already there for AI. With generative AI, though, it takes it to another level. Beyond elections, there's pretty substantial concern about like the fake porn, about the use of especially image generators to create non-consexual, sexual content. Last month, there was kind of an uproar about Taylor Swift being the victim of this, that there had been a channel on 4chan that was treating the creation of deep fake porn, featuring celebrities as kind of a game. You know, and I looked at this thread and there are a lot of examples and a lot of them are really realistic and, you know, it doesn't just affect celebrities. There have been cases where like high school and middle school students have been the subject of this sort of thing. Oh my god. Yeah. You know, a lot of times this sort of thing is used for revenge and the fact that this technology is now so easy to use and so easy to access is deeply concerning for a lot of authorities. You know, that said, many of the AI companies that make these tools are taking steps to not have their tools respond to prompts asking for images of a sexualized nature or prompts that ask for an image of a real person or a public figure, that sort of thing. But you know, it's as that 4chan thread show, it's pretty easy to work around those restrictions. What's one thing you're watching, you know, that we all should be paying attention to as the campaigns ramp up in the election approaches? I am personally really interested in the Liars dividend phenomenon. That's the ability to so doubt by claiming that authentic evidence is fake. This was already manifested in the Israel Hamas war. A lot of people were saying it's not so much that AI generated content is convincing people of one particular view. It's that the fact that people can create AI generated content means that people don't trust anything that they're seeing about the conflict. And my concern is that somehow that's going to play a role in the mega cycle of elections around the world. Because really all you need is a well-timed deep-fake audio file to land in October. And then you have an October surprise that is very, very hard to disprove one way or the other. And it's very hard to track who created it. So you don't have the kind of accountability measures in place at least at the moment. Yeah, I mean, unfortunately it seems like it can lead to everyone's credibility just for existing in the same world as this stuff. Yeah, there was a letter that Michigan Secretary of State, Jossam Benson, wrote in October. She sent it to Senator Chuck Schumer of New York. And she said that AI generated content may supercharge the believability of highly localized misinformation. And I think in an election in the US where like particular precincts might determine how the race goes. That's pretty terrifying. I mean, she said those seeking to sway outcomes are so chaos, manless AI tools to mislead voters about wait times, closures or even violence at specific polling locations. And the fact that this technology could be used to prevent the democratic process from proceeding as it should is something that everyone should be watching and then trying to prevent, though it's unclear how exactly that should take place. Tiffany Shu, thank you so much for joining us today on the weeds. Yes, thanks for having me. So this brings us back to our original question. On a scale from one to ten, how worried should we be about AI this election year? I'm going to put a set of seven. Not only has the technology gotten better, but this tech now offers plausible deniability for anything out there that is true. I realize that isn't the most satisfying answer. If I'm honest, it's actually a really unsettling one. But that just means we're going to have to be really diligent and pay attention to where we're getting our information from. Do you have a question? It can be about anything. The election, the economy, how something works or how to solve a problem you see in the world. You ask it and we'll find answers and explore ideas. Email your questions to [email protected]. That's all for us today. Thank you to Tiffany Shu for joining me. Our producer Sophie LaLonde, Krishniyalla engineered this episode, Melissa Hirsch fact-checked it, A.M. Hall asked today's big question and they're also our editorial director. And I'm your host, John Golan Hill. This podcast is part of Vox, which doesn't have a paywall. Help us keep it that way by going to vox.com/give.