Prediction Market Betting Is Getting People Banned and Arrested

back! This week, senior writer Kate Knibbs joins Brian Barrett, Zoë Schiffer, and Leah Feiger to discuss the two incidents that have been making noise in prediction markets: George Santos getting lifetime ban from Kalshi and a Google engineer’s Polymarket insider-trading case. Plus, reporters reverse-engineered Flock’s AI-powered person-search tool and spoke to experts about the accuracy of such technology, and the hosts unpack the online meltdown over how to talk about “rogue” AI agents.

Articles mentioned in this episode:

• George Santos Just Got Hit With Kalshi’s First-Ever Lifetime Ban
• Google Engineer Accused of Polymarket Insider Trading Says He Was Just Gambling
• This Is Flock’s AI Search Tool for Cops
• What We Still Don’t Know About OpenAI’s Hugging Face Hack
• The Logical End Point of AI Job Interviews Is Two Bots Talking to Each Other

You can follow Brian Barrett on Bluesky at @brbarrett, Zoë Schiffer on Bluesky at @zoeschiffer, Leah Feiger on Bluesky at @leahfeiger, and Kate Knibbs on Bluesky at @knibbs. Write to us at [email protected].

How to Listen

You can always listen to this week's podcast through the audio player on this page, but if you want to subscribe for free to get every episode, here's how:

If you're on an iPhone or iPad, open the app called Podcasts, or just tap this link. You can also download an app like Overcast or Pocket Casts and search for “uncanny valley.” We’re on Spotify too.

Transcript

Note: This is an automated transcript, which may contain errors.

Leah Feiger: Hey guys, so nice to be back after our two-week summer break.

Zoë Schiffer: Longer for me. It's so nice to see you guys.

Leah Feiger: I missed you.

Zoë Schiffer: I missed you.

Leah Feiger: Brian, don't worry. We missed you too.

Brian Barrett: I assumed that the you was for all of us.

Zoë Schiffer: It was the royal you.

Brian Barrett: But now that you've carved me out, that's telling. Thank you.

Zoë Schiffer: Welcome to WIRED's Uncanny Valley. I'm Zoë Schiffer, contributing editor.

Brian Barrett: I'm Brian Barrett, executive editor.

Leah Feiger: And I'm Leah Feiger, director of politics and science. This week we're discussing some recent high profile incidents in prediction markets. Former US representative George Santos was the first person to be given a lifetime ban from Kalshi, and a Google engineer was accused of insider trading at Polymarket. We'll get into why both of these cases show that there are still plenty of blurry rules when it comes to prediction markets.

Brian Barrett: We'll also check in on perhaps the most hated tech in the US at the moment, maybe a runner-up to data centers, Flock surveillance cameras. This week, WIRED reporters were able to recreate Flock's AI powered search tool, the same one that has a track record of being misused by police officers. We're going to break down how it works, and whether the backlash against Flock cameras might have reached its boiling point.

Zoë Schiffer: And because I simply cannot help myself, we will also be getting into why tech bros on X are melting down over how to discuss rogue AI agents and what happens when they jump their guardrails and coordinate attacks on other systems.

Leah Feiger: All right. Do you guys remember George Santos?

Brian Barrett: How could I forget?

Zoë Schiffer: Absolutely do, but wow, did not expect we would be talking about him again this soon, I have to say.

Leah Feiger: Former US representative expelled from Congress in 2023, accused of fraud and theft, and doing it with such a plum that the story just couldn't help but take national news by storm.

Brian Barrett: I want to be clear too. He wasn't just accused, he went to jail.

Zoë Schiffer: Yeah, I was going to say very credibly accused.

Leah Feiger: Anyways, he's back in the headlines this week. On Monday we learned that Santos was hit with a lifetime ban from prediction market platform Kalshi, and a $71,000 fine, allegedly for trying to manipulate a bet on whether he'd show up to Trump's State of the Union. To tell us all the full details, we are joined by WIRED senior correspondent Kate Knibbs.

Kate Knibbs: Hi guys, how are you doing?

Leah Feiger: Kate, so happy that you're here. Can you take us through my favorite story of the week?

Kate Knibbs: I would love to tell you guys about this. And I will say, I feel similarly like Santos, yes, he does do crimes but he does them with such flare that it's hard to hate him.

Zoë Schiffer: It's a bad thing that's very fun to watch.

Kate Knibbs: Yes. So Santos, at this point when this all begins, he's already expelled from Congress, already gone to jail. He's out of jail, he's making Cameo videos for money, he's doing his thing. So, he posts that he will be attending the State of the Union. And then, while the State of the Union is going on, he posts that he's actually stuck at the airport and he's not attending it. And then he says, "FML." There was a market on Kalshi about whether Santos was going to attend the State of the Union, and someone made a bunch of money off of it. And Kalshi quickly figured out that it was him because Kalshi has know your customer requirements. His identity was not a secret to them, it was not hard for them to check and see that the person who made money was George Santos.

Leah Feiger: Awesome.

Kate Knibbs: And it's against their rules to place trades or wagers on events that you're a participant in. It's like market manipulation, insider trading, whatever. So even before this happened this week, the CFTC actually also knew about this because Kalshi told them, and they already fined him money for this specific incident. What happened this week was apparently there was this back and forth between Santos and Kalshi, and they fined him $71,000 and then weren't going to lifetime ban him. The same day that they announced the Santos thing, they announced a few other politicians who got in trouble for betting on themselves, and those politicians got much smaller fines and temporary bans.

Zoë Schiffer: Wait, do we know how much money he made on the market?

Kate Knibbs: I believe it was around 17 grand. I have to double check, but it was in the tens. They ended up enacting this ban basically because he was being a little bit of a prickly pear to work with and wasn't playing ball. So, this is Kalshi basically deciding that they hate George Santos, which is funny. He also immediately, he wouldn't give us comment, which was annoying, but he posted on X some kind of bitchy things about Kalshi. He immediately was like, "Let's see how long you guys last," kissy emoji. So, he doesn't seem super contrite, and I honestly don't even know if he's going to pay Kalshi. I asked them what they would do if he doesn't pay and they said that they would pursue legal action, and I am going to be following that because it will be interesting to see.

Zoë Schiffer: This is my big question. Does George Santos want to get away with these crimes, or is there something else going on here? Because it's one thing to participate in the market that you are literally a participant in in real life, but it's another to be blatantly posting about it in very obvious ways on social media.

Kate Knibbs: Well, if he did do it basically on purpose, he wouldn't be the first politician to do so. There's actually already a storied history of politicians doing insider trading on Kalshi, getting caught and then saying it was a bit. There's two other people who have already done this. Kyle Langford, who was caught this spring, and then this candidate, Mark Moran, who said that he did it on purpose as a sort of performance art piece against prediction markets.

Brian Barrett: This is my new get out of trouble card just in general, that I was just doing a bit. So, if I'm ever slow to put an edit back to you guys or anything like that, it's a bit. It's for my art.

Zoë Schiffer: And we totally get that.

Leah Feiger: I love it. I'm sorry, I think that there have just been so few rules and regulations, as you've reported, Kate, and there's just been a lot of gray areas that people have been tap dancing around that I kind of love an example where Kalshi had absolutely no choice other than to say, "Every single part of this was so wrong. Here are the facts from A to Z. This is it. This is the line in the sand." And I think that having examples like this is only good. So, that's where I'm coming from, I suppose.

Kate Knibbs: I get that. And you know what has been interesting? Because I'm so steeped in this prediction market world, sometimes I forget that other people don't know all of the arcane rules. And a lot of people have reached out to me about this story and been like, "Why is Kalshi allowed to fine him? That's weird. You don't get fined by FanDuel if you do something." And so, it's been an opportunity to explain to people it's because Kalshi is a federally regulated financial platform. It actually is imbued with the power to find people in the same way that the CBOE is. And that is something that most people didn't know, and now maybe they do, and we can thank George Santos.

Leah Feiger: Can we talk about Polymarket and the differences here? I referenced in the introduction the Google engineer who was arrested by US authorities back in May for alleged insider trading on Polymarket. Talk to me about this, Kate. How is this story different? How is it similar? How is Polymarket and Kalshi at this point different and similar?

Kate Knibbs: The Google engineer's case is different in a bunch of different ways. He made a lot more money, he made over a million dollars. So, the scale of it is quite impressive. He got arrested, Santos didn't get arrested for this. He's just gotten fined. He got arrested and he is pleading not guilty and is attempting to say that he's not guilty because he was actually just doing betting. And because he was just doing betting, the federal government has no grounds to attempt to try him under laws that pertain to commodities violations. Basically, the Google engineer was in Europe. He's a European living in Europe. When he made these trades using information that he learned because he was employed by Google, he knew who the most searched person of 2025 was going to be. He allegedly knew. And he's not, in the court filing so far, he's not trying to say, "I didn't do it." He's trying to say, "If I did it wouldn't be a crime as you're saying it is, because I was just doing international betting and what does that have to do with commodities law in the US? How was I insider trading? These aren't trades, they're just wagers." So, it's a really interesting case because he's basically allying himself with all of these state regulators that are currently battling the federal government and Kalshi and Polymarket over whether they're actually offering bets or trades, and it will be very interesting to see how it plays out. And the argument that he's making is also being made by the other person who was arrested for insider trading, the special forces operative. He's also making this argument. The two guys, I don't know if their lawyers are talking or what, but they're both coming out and saying, "Well, whether or not I did it, it doesn't really matter because this isn't trading, it's just gambling. And why are you trying to pretend that you can insider trade on a gambling platform?" And so, yeah.

Leah Feiger: Kate, thank you so much for joining us. I have no doubt that your reporting on how all of this continues, especially with the midterms, will be invaluable.

Brian Barrett: Thank you, Kate.

Kate Knibbs: Anytime.

Brian Barrett: OK, we're going to switch gears. We're going to talk about Flock cameras, which have been pretty reliably in the headlines the past few weeks, and usually not for great reasons. Lots of cases of police officers allegedly abusing the systems to track their colleagues, to track exes, and a lot of protests around them, a lot of people canceling their contracts with them, municipalities. This week, WIRED reporters Dhruv Mehrotra and Dell Cameron got their hands on the code behind how Flock's new AI-powered search tool works. This is actually some fun wizardry, they were able to take the code and recreate what it looks like. So, they're able to sort of interact with the Flock system in the same way that a police officer would. They don't have access to all of the backend stuff. And here's the thing, when you all think of Flock cameras, what do you think of them doing?

Leah Feiger: Surveillance.

Brian Barrett: But what kind? What specifically do Flock cameras look at?

Leah Feiger: License plates?

Brian Barrett: Yeah, they're license plate readers. Yes, this tool is not a license plate reader, it allows you to look up people. You can type in a general description. You can say, "Someone wearing scrubs, someone who has tattoos." You draw a little shape on a map, a little circle on a map, and the system will run a continuous automated search for a match across every camera inside of that circle.

Leah Feiger: That's so bad.

Brian Barrett: It's pretty bad. It's not great. And I think it's a capability that most people don't assume Flock has or offers, and it's a relatively new system, but we do know it has been deployed in some places already so it's out there. And not only that, there's not really a moderation framework in place. Flock recently announced some ways to make sure that people aren't abusing the license plate readers. This seems like so open-ended and such a chance to look for whoever you want, wherever you want using all of these cameras. It's really kind of stunning.

Zoë Schiffer: I want to dig in a little more on the it's really bad thing, because I mean, I think a lot of people are in agreement on that, but can you, Brian, just make the case for why this technology would exist? What is Flock's argument for why it could be good, and then why do you think it's bad?

Brian Barrett: So, I think the argument for why police officers like it is it's another tool to identify people who were near the scene of a crime, or suspicious looking people or whatever. So, I think it's just the more information, I would assume, that the more information you have as a police officer, the better you feel about the investigation. In terms of why it's bad, I think a few reasons. It is not a reliable tool necessarily. I think there's a lot about these sort of person recognition tools that aren't necessarily 100 percent you're going to get it right every time—

Zoë Schiffer: Like it can misidentify people.

Brian Barrett: Yeah, but even without that I think we already live in such a surveillance sort of state that we've become inured to it, but I don't think we should. I think the idea that anywhere you're walking you can be identified and brought into some sort of police probe because you happen to have tattoos, or happen to be wearing scrubs, or whatever other physical identifier feels like a pretty serious violation of your rights.

Zoë Schiffer: I will say it's been interesting talking to people about it because you really do get a lot of people who live in San Francisco being like, "Petty crime was a huge problem and now it's less of a problem," and that's really nice. And then there's a lot of people who are much more concerned about civil liberties violations.

Leah Feiger: Very specifically with the Flock cameras, the way that I've been talking about it with people is these cameras aren't operating individually. We're still relying on police forces to be using them and trolling their own databases, which then implies an inherent trust of the police state in addition to the surveillance state, which is already mired in a variety of reliability issues, accountability issues. So, the idea that a police officer in wherever, in San Francisco can be "fighting crime with Flock cameras" feels like incredibly surface level like, "OK, go for it, Batman." But when you're also like, "Well, which of these police officers have received reprimands from their bosses? Which of these are actually using these services to look up ex-girlfriends or family members?" The idea that this technology, which on its face doesn't have to be entirely negative, but is controlled by certain forces that could turn it to be so is very, very concerning to me.

Brian Barrett: It reminds me a lot of, or at least rhymes with geofence warrants, which for those who don't know, a geofence warrant is where law enforcement could go to Google and say, "Hey, I want to know everything about all the cell phone data, location data for anyone who was near a particular place during a particular time." This was, I think used, if I remember correctly, pretty extensively around the January 6 riots. They found a lot of people through these geofence warrants because they were able to say, "Hey, Google, who had a phone inside the Capitol at this time who was not a member of Congress?" The Supreme Court just a couple months ago ruled that those kind of geofence warrants do count as a search under the Fourth Amendment. So, you need a reasonable search requirement, it's a higher bar to actually conduct these things than there was because they were just going willy-nilly. Again, not a lawyer, but as a simple country tech editor, it does feel comparable in the sense that if you're going to be searching who is in a certain area at a certain time based only on physical characteristics or whatever else, there should be a higher bar for being able to do that. So, it is an unsettled question for sure, like what amount of surveillance is worth the return, but I does feel like people are leaning towards this world in which we've gone too far with Flock in a bipartisan way. We've had Florida just banned plate readers from state highways. Texas has stopped state funding for Flock cameras. I live in Alabama. Alabama, of all places, has started to say, "We need to take another look at these Flock cameras." They haven't done anything yet, but they're talking about it. And that's something to the point of like, "Yeah, we need to figure out where this lands." I think we're in that process right now, and it's landing on the side of maybe a little less Flock in everybody's lives.

Zoë Schiffer: I'm going to drag us to an even darker place on the internet. This past week, a blog post by the tech podcaster, Dwarkesh Patel, went viral on X. It had Silicon Valley elites up in arms. Basically, it was a summary of the OpenAI Hugging Face hacking incident earlier this summer. And to refresh everyone's memory, even though this feels like old news, this was when two OpenAI models went rogue and hacked the AI platform Hugging Face during a routine security test. Patel's way of framing the incident is what drew a lot of attention and ire online. He described the hack as part of this larger operation run by a group of AI agents, or as he called them, AI civilizations that rose and fell inside OpenAI during these training runs.

Leah Feiger: Civilizations.

Zoë Schiffer: I know, I know. Just stay with me.

Leah Feiger: Age of Empires, folks.

Zoë Schiffer: He dug into how the bots worked together to complete impossible tasks that they had been assigned. He nicknamed them based on their "behavior." One was called Alexander the Great, for example. He says there were points where you could see the bots sacrificing themselves for the good of the whole. All in all, the post had a ton of anthropomorphization. It reads like something that was written by a guy who loves to talk about the Peloponnesian War, which in fact, I feel like a lot of these tech podcasters do love to talk about that. But after the post came out, it was interesting. You really saw two takes, maybe three across X. On the one hand, you had people like Kevin Roose and Matthew Prince being like, "This is really scary and we should pay attention to this." And then on the other side, you had Tribhuvan Krishnan being like, "This post uses way too much anthropomorphic language." And then you had people like Chamath Palihapitiya who was like, "I think this whole thing is probably a psyop to shut down open source AI development." I made you guys read it, what did you think?

Brian Barrett: Well, I read this stuff also sometimes. So, I think we need to go to Leah for the pure unadulterated, "I am reading this only at gunpoint," take.

Leah Feiger: I thought it was a joke. I started reading it and I was like, "Is this Katie Weaver writing? Is this fan fiction?" I was sort of shocked at the idea of, to Krishnan's point, the anthropomorphization, say that five times fast, of these bots as a community. And perhaps the English language really doesn't have, despite our many decades of science fiction literature, shout out Tim Marchman who is currently listening and going like, "No, we have the language," but I don't know if we actually have the language to talk about what's happening here because they're not making communities. These are not synapses firing. This is very specific, something different. So, that was very odd for me. But also the idea of where this could go in the future was quite scary, because the idea that someone was not only recounting this, but was kind of pumped about it. It was spoken about lovingly, I felt. Is that incorrect?

Brian Barrett: Well, I don't know. I can't look into another man's heart, but I do think, Zoë, is it fair to say what struck me about this in some ways is A, two things. One, I think both can be true that it is over anthropomorphized. You said that so well earlier, Zoë, you made it look so easy. But then also, yes, it is scary that these bots acted this way. I think objectively it is sobering that these AI agents can do this. But also I feel like some of these people are trying to have it both ways. I think when AI is going to be your boon companion and your best friend and take over the world, it's fine to talk about them in human terms, but if you're talking negatively about them, "How dare you? They're just a bot, you're dramatizing it. Why are you being so ridiculous?" You know what I mean? I think it's really using language, as everyone has since beginning of time, but as is convenient for your case currently, how you talk about bots and AI agents really changes on what you want to get out of it.

Zoë Schiffer: Yeah. I thought it was interesting because I feel like at this point we've all read a lot about this particular hack. And on one hand, purely on a literary level, I was like, "This thing is fun to read." He really did a good job being like, "Now, this battalion of bots did this crazy thing." On the other hand, it reminded me just as a writer how important it is to really think about the language we're using, because it did feel like rather than the focus of that article being what actually happened, the focus very much became how we talk about what happened. I also think it is concerning to see people like Chamath Palihapitiya, who is the host of the “All-In” podcast, but also has been involved to some degree, unofficially involved in the Trump administration and going to the White House and certain things, basically say that any level of investigation or media attention on a security incident is a psyop. I was just like, "Ooh, that seems like a really scary place to be going where you're so pro AI development that any publicizing of a security incident is actually proof that forces that want to slow down AI development are secretly seeding these messages places."

Brian Barrett: Yeah, I think that's a really good point. It kicked off a healthy debate, but not the right debate. And people are using it to distract from, I think, again, a really serious finding. I'm glad that it got more attention than it had been getting and that Dwarkesh really took the time to dig into it. But yeah, it's infighting over lexicography or whatever is not really where we need to be right now.

Leah Feiger: Coming up after the break, we're sharing our WIRED/TIRED picks for the week. Stay with us.

Zoë Schiffer: OK guys, it's time for our WIRED/TIRED segment. Whatever is new and cool is wired, whatever passe thing we're completely over is tired. Who wants to go first? Just kidding. I'm going to call on Brian. Brian, you're going first.

Brian Barrett: Thank you, Zoë. That's great, I win. So, my WIRED is AI job interviews when you go to interview for job and it's AI on the other end of it. This is an increasingly common thing. My WIRED is creating your own digital twin to do the interview for you on your end. Kate Taylor, who's a senior writer on our business desk, talked to a guy recently who, I found this amazing. He had talked to the same AI chatbot for different jobs five times because it was from a centralized recruiting agency. So he would apply to a job and it would be the same ... This AI chatbot called Riley, and you'd have to talk to Riley about why he was right for a job. Never heard from a human ever. So he said, "You know what? I'm just going to make my own digital twin of myself. They'll do the interview—"

Leah Feiger: Good stuff.

Brian Barrett: Did that a couple times. Then he made an AI chatbot that he thought was the idealized candidate and had that do an interview. None of them ever got called back either. I do recommend going to find cases around WIRED because we did publish video of this guy's chatbot talking to the interview chatbot, and it is delightful.

Zoë Schiffer: That's so great. OK, Leah, what about you?

Leah Feiger: My WIRED and TIRED this week is incredibly bad, and it's clear that I'm slightly phoning it in, but all I can think about right now, you guys, is the midterms are so close. That's my WIRED because I'm so excited.

Zoë Schiffer: You are so tired though. You are straight up so tired.

Leah Feiger: I'm so tired, but I'm wired because the midterms are pretty much here. Primary season was a slog, and it's done, and I'm just so excited for what's to come. My TIRED is some Democrats are starting to count their chickens before they hatch, and that is wild to see. Our wonderful senior correspondent and author of the Inner Loop newsletter for WIRED, Hugo Lowell, has a piece out all about how Democrats are already planning for subpoenas to fly once Democrats take control of the House, and what repercussions they could land on the administration should they not comply. And while that's all very lovely and good, the elections have not happened yet, and there is so much disarray within the Democratic Party at the moment in some very key states, I'm looking at you, Maine, I'm looking at you, Michigan, that I'm not ... I don't know. It's a little bit concerning to watch that happen from afar.

Zoë Schiffer: I mean, I just have to say, I feel like if Democrats learned anything over the past eight years, it's to not count the chickens. Don't count the ... Just wait. Don't count them.

Brian Barrett: Don't count them.

Zoë Schiffer: Maybe even after they hatch—

Brian Barrett: Don't do it.

Zoë Schiffer: The chickens could be really old and tired, and that could be a whole thing. My WIRED is combining an AI agent with your savvy human friend to plan your life. Leah's dying right now as I say it. Hear me out. I'm going to Venice next week, humble brag, brag bag. Don't know a lot about it, asked Leah for recommendations. She gave me a bunch, I was so excited. Didn't want to book all of this myself because I had a very busy week. Tried out one of the buzzy AI assistants that is taking over Silicon Valley. That's one of the ones you just text and say, "Hey, can you do this thing?" Speaking of surveillance, you do have to give it access to a lot of stuff, so that's that. And thus, I'm not going to talk about which one I'm using lest my phone get hacked. However, it booked all of my reservations for me, and I did kind of have a moment where I was like, "This is sick." I just gave it the list and was like, "Based on where I'm going these days, can you make the reservations, lunch and dinner?" Done.

Leah Feiger: Wait, but how? These were mom and pop middle ... They have a phone number that they barely answered.

Zoë Schiffer: At times, it emailed them or WhatsApped them.

Leah Feiger: Oh …

Zoë Schiffer: I know.

Leah Feiger: ... my God. Wait, this is sort of transformational. Usually when we go places I make a list and give it to Rodrigo, and I'm like, "And now you have to spend the next five days figuring out between siesta and dinner when you're able to call."

Brian Barrett: So it's sort of like having a Rodrigo, but a digital twin.

Zoë Schiffer: Rodrigo's one of our most avid listeners. Hi, Rodrigo.

Leah Feiger: I'm actually really excited to hear how this goes.

Zoë Schiffer: If I show up to the restaurant and they're like, "We've never heard of you," that would be really funny. No, but I feel like, I mean, kind of like the surveillance conversation, it is this question of you need to give agents enough access to your systems in order for them to be really useful, but then we just don't have the right tools in place to stop them if they get very motivated from going rogue. That was the whole thing about the OpenAI incident is it gave them an impossible task and then they really tried to succeed. So yeah, I think that there are huge risks with these systems, but again, a lot of people in Silicon Valley are using them, me included apparently. That's our show for today. We'll link to all the stories we spoke about in the show notes. Adriana Tapia produced this episode. It was mixed by Pran Bandi, who's also our New York studio engineer. It was fact checked by Matt Giles and Daniel Roman. Marc Leyda is our San Francisco studio engineer. Kate Osborn is our executive producer, and Katie Drummond is WIRED's global editorial director.