DHS’ Secretive ‘Predictive Policing’ Unit Pulling People Over

9 Sep 2026 · 43 min · 16 chapters

Ask about this episode

Ask anything about it. ChatGPT or Claude reads this page and answers with the times it was said.

Connect VO and ask about every podcast you hear, including the moments you saved. Add to ChatGPT · Add to Claude

In short

404 Media’s Uncanny Valley episode discusses DHS/Border Patrol “Predictive Intelligence Targeting Teams” (PITT) that use law-enforcement-sensitive databases to flag Americans based on financial activity patterns associated with illicit narcotics, then relay names/vehicle indicators to local police as a pretext for traffic stops. It also covers “adversarial fashion”: a digital camouflage shirt meant to confuse YOLO object-detection surveillance cameras.

Guests

Host Joseph with 404 Media co-founders Sam Cole and Jason Kepler. No other guests are interviewed.

Key claims

PITT is secretive; “predictive” targeting is used without probable cause or a specific suspected crime. A DHS memo of investigation tied one stop to Border Patrol analyst Matthew Phelps and a Spokane-sector targeting analyst. Another PITT team was inferred via court records/texts and a WhatsApp profile (“tech ops PITT”).

Notable examples

Kyle Olson (California cannabis worker driving to Wisconsin) was stopped in Montana for an “obstructed license plate,” then charged with DUI and possession with intent to supply after DHS targeting intelligence. Fashion segment: Simon Weckert’s gradient-pattern shirt defeats YOLO person detection in a studio test; it went viral and sparked debate about surveillance and algorithm improvement.

Written by AI. May contain mistakes. Listen to the episode to check what was said.

Chapters

Tap a time to open that second in VO

Hosts Introduction

0:27 to 1:41

Hosts Joseph, Sam, and Jason introduce themselves and hint at upcoming content.

“Listen to new episodes every Thursday, wherever you get your podcasts.”

Live Podcast Announcement

1:41 to 1:55

Announcement of a live podcast event with The Intercept in LA.

“Jason, do you want to take us through this first story?”

Introduction to the DHS Story

1:55 to 2:16

Joseph introduces a critical story about DHS's predictive policing unit.

“Yeah, quickly, if you are in LA, we're doing a live podcast with The Intercept on Thursday at The Loved One at 7.30pm.”

Understanding DHS Predictive Intelligence Teams

2:16 to 3:40

Discussion on how DHS analyzes financial data to flag individuals for stops.

“A Secret of DHS Predictive Policing Unit is Analyzing Americans' Financial Habits and Pulling Them Over.”

Kyle Olson's Experience with DHS

3:40 to 4:32

Narration of Kyle Olson's encounter with law enforcement based on DHS data.

“And these are called Predictive Intelligence Targeting Teams, or PITT.”

Details of the Traffic Stop Incident

4:32 to 10:01

In-depth account of the traffic stop involving Olson and the justification used by police.

“Yeah, so Kyle Olson was driving from, I think, California to Wisconsin.”

Implications of Predictive Policing

10:01 to 12:18

Discussion on the broader implications of DHS's surveillance and predictive practices.

“but then I gave it to the cop and they pulled them over.”

Critique of Predictive Policing Methods

12:18 to 14:00

Critique of the ethics and legality surrounding predictive policing and its practices.

“You pull people over because they have committed or are suspected of committed a specific offense, right?”

Exploring Predictive Policing and Its Implications

14:00 to 23:30

Discusses the concept of predictive policing and its controversial use by DHS.

“Like it was, he had some amount of weed.”

Exploring Predictive Policing and Its Implications

25:10 to 26:35

Discusses the concept of predictive policing and its controversial use by DHS.

“Either you get to hear what you're listening to, or you get to hear what's happening around you.”
Show all 16 chapters

Exploring Predictive Policing and Its Implications

27:33 to 28:39

Discusses the concept of predictive policing and its controversial use by DHS.

“So for fun, I googled myself last week and found my home address, my phone number, and a relative I'm fairly sure I've never met all on directory sites I never signed up for.”

Digital Camouflage Shirt Overview

28:59 to 30:00

Discussion about a shirt designed to confuse AI surveillance cameras.

“Sam, do you want to take this one, the Jason ring?”

The Mechanics of Adversarial Fashion

30:00 to 38:08

Exploring how the shirt works against AI recognition and its implications.

“So essentially this shirt is, it's like made up of a bunch of blobs and we can talk about how it was made in a second, but it's like green and black and pink and red.”

The Broader Conversation on Surveillance

38:08 to 42:00

Discussing the societal implications of surveillance and the role of art.

“It's the point like to provoke a conversation, a response, et cetera.”

Celebrating the Third Anniversary

42:00 to 42:33

Learn about the activities and events marking the podcast's third anniversary.

“we're going to talk about our third year anniversary.”

Celebrating the Third Anniversary

42:36 to 42:46

Learn about the activities and events marking the podcast's third anniversary.

“You'll get unlimited access to our articles and an ad-free version of this podcast.”
Hear the part that matters, and keep it.Open this episode in VO. Double tap your headphones to save a moment as you listen.
Get VO free

Transcript

Automatic transcript. May contain errors.

0:00The Wired newsroom is known for award-winning reporting on how technology shapes our world. On Wired's Uncanny Valley, we take that curiosity even further. Each week, journalists from Wired break down the biggest stories in tech while speaking directly with the people building, challenging, and reshaping the future. Is the AI boom sustainable? How do you protect your privacy in an age of constant surveillance? Uncanny Valley tackles the questions driving today's tech debates and lighting up your group chats. Listen to new episodes every Thursday, wherever you get your podcasts.

0:57and this found the company and needs your support. To subscribe, go to 404media.co. As well as bonus content every single week, subscribers also get access to additional episodes where we respond to the best comments. Gain access to that content at 404media.co. Also, do remember to subscribe to our YouTube channel where you can watch all of our episodes, including this one. Subscribe at youtube.com slash at 404media.co. I'm your host Joseph and with me are two of the other 404 Media co-founders. The first being Sam Cole. Hey. And the other being Jason Kepler. Hello, hello. We survived our party just about.

1:40We will talk about that a bunch more in the subscribers section at the end of this episode. But it was a lot of fun. Let's save the chat until then. Jason, do you want to take us through this first story? Yeah, quickly, if you are in LA, we're doing a live podcast with The Intercept on Thursday at The Loved One at 7.30pm. We'll drop a link to RSVP for that. I think there's only a few spots left. So if you want to come to that, please do. All right, let's talk about this story that you did, Joseph. It's called, A Secret of DHS Predictive Policing Unit is Analyzing Americans' Financial Habits and Pulling Them Over.

2:23That's a bit of a long headline. It's a good one, though, because there's a lot going on here. So I guess we'll get to the specifics, but what is the new, what is this? What's going on here? Yeah, the new thing here is that DHS, or more specifically Border Patrol, is running these secretive predictive policing units that analyze Americans' financial activity and other data, sends that information to local cops, or rather really the names of these people. Local cops then find a pretense to pull them over. They're not suspected of any sort of particular crime. You know, there's not probable cause here.

3:06They're not going after somebody because, oh, they trafficked XYZ amount of drugs or something like that. They pull them over, not suspected of a particular crime, because DHS thinks they're probably worth searching. And they probably, maybe, have stuff on them that would be of interest to DHS. So there's all of that stuff. And one of the really big things that we've revealed in this article is simply the name of this unit, or rather units, and we'll explain that in a bit. And these are called Predictive Intelligence Targeting Teams, or PITT. I guess we can say PITT for short. The thing is... It's a different PITT.

3:52Yeah, yeah, yeah. The thing is, it's not like I can really go to, I don't know, a DHS podcast and see how they pronounce it. Do they say PIT? Do they say P-I-T-T? Because when you Google, or rather before we publish this article, when you Googled predictive intelligence targeting team, literally nothing came up. And I thought that was very, very interesting. and one of the reasons why we're revealing the name of this predictive policing unit. Yeah, so this again, no one knew what they were doing here, but this came to you through a guy named Kyle Olson. Who is he and what happened to him? Yeah, so Kyle Olson was driving from, I think, California to Wisconsin.

4:45he works on a cannabis farm in California and he was taking some edibles, some oil, some marijuana back to Wisconsin. Right. And I'm not an expert in all of the state laws around marijuana and cannabis. What I'll say is that he believes that what he was doing was legal. I also think it's neither here nor there, right? Because it's much more about DHS didn't have probable cause or suspect of a particular crime, that sort of thing. And we'll get to that. But he was driving, as I said, from California to Wisconsin. He has his dog in the back. He has a black SUV. there's some luggage racks on the back that's sort of attached to the rear bumper and then eventually when he's crossing through montana i believe that is when a state police officer pulls him over so he's pulled over and uh like why why did he think they were stopping him or Like, what did, what originally happened here?

6:06Olsen sent me a bunch of body cam footage from the police and also sort of his dash cam footage, you know, a camera in his own SUV. And that includes a rear camera. And I watched that. And you see the cop car, the Montana, I think, Highway State Patrol. They pull out onto the interstate. They follow Olsen for a while. Eventually, they put their lights on. They pull Olsen over. And then in the body cam, this sergeant, Sergeant Beck, I think, goes up to the side of Olsen's car and he says, Hey, nothing major. We're just pulling you over because your license plate is a little bit obstructed. The sergeant thinks there's a cover on the license plate, like a glass cover sort of thing that's shielding it.

7:03He also thinks that those big luggage racks might be obstructing it as well. But crucially, that is the given pretense and the given reason for the stop. And Sergeant Beck says it quite explicitly, like, hey, just so you know, this is the reason we're pulling you over is because you have an obstructed license plate. Right, but that's not why he was pulled over. Like, as you later found out, the entire reason we're doing this story is he was flagged by the pit, this predictive intelligence team. So how did he find out about it? And then how did you kind of, what gaps were you able to fill in? Yeah, so Olsen's case has progressed, right?

7:50He has been charged. He's been charged with a DUI because they took his blood at the time. and I think he was allegedly under the influence of cannabis, so I'd have to double check that. But he was charged with a DUI and then later charged with possession with intent to supply because of the edibles and the marijuana and stuff in his car. So that case has been going on and as part of that, he's been given discovery, right? Which is, hey, here's the evidence that the state or the authorities have against you. And in that, he was given a very, very interesting DHS written document. And it actually lays out the real reason for this stop, which is that Border Patrol is running this predictive intelligence targeting team.

8:41An officer, a Border Patrol agent from one of those teams in the sort of Washington state sector, was going through Americans' financial data. and other information. They identified Olsen as a potential person who was engaging in narcotics trafficking. They then took that information and then they provided it to Sergeant Beck, specifically the cop who then pulled Olsen over. So in other words, the license plate stop was a pretense. It was just a reason to get him over at the time. The real reason that Olsen was stopped was because of a DHS predictive policing unit that is analyzing all of this data, pinpointing vehicles and people and thinking, hmm, they look interesting.

9:34We should probably pull them over. And I don't know, it's pretty rare for us, I think, to see a document like this. We'll talk about other really, really crucial reporting on DHS predictive policing in a moment. But having a document like this, DHS says, yes, I'm analyzing the data, and then I thought this person was doing this. They're not suspected of a crime, but then I gave it to the cop and they pulled them over. I mean, it's a really, really crazy document. When I... Because the turnaround of this article was pretty quick, right? I mean, that's the benefit of how we operate. We can move quite quickly.

10:13We're still very careful and diligent. We don't have a ton of bureaucracy, right? So we can do that. I got this document who, to be clear, Olson gave it to me. I then showed it to you, Jason. What did you think when I showed you this document? Well, I think let's tell people specifically what it says because it's called a memorandum of investigation. And so it's a DHS document that explains a bit about the investigation in here. And it's from someone in the Spokane sector. of Border Patrol Targeting and Intelligence Division, which I guess oversees this PITT group. And he wrote that he reviews, quote, law enforcement sensitive databases and then uses that to develop intelligence that they hand over to local law enforcement, which you've said.

11:07But it specifically says that, quote, information contained within law enforcement sensitive systems suggesting financial activity patterns commonly associated with illicit narcotics activity. And then they basically asked them to like use, to interdict this person. So pull them over. Yeah. Then in the actual like report detailing the stop, Beck wrote, quote, it was relayed to me by U.S. Border Patrol targeting analyst Matthew Phelps that he had identified a vehicle that had indicators of possible narcotics trafficking. So this again, it's like DHS is using unknown, like we don't know which government databases, we don't know which financial information to tell local cops like what to pull people over for because they think that they may have possibly committed a crime.

12:06That is really wild. It's just a really wild thing to do. That's not how typically the justice system works in America, right? Like, you don't pull people over on a hunch. You pull people over because they have committed or are suspected of committed a specific offense, right? And of course, you know, it's very convenient in this case that the license plate did look to be obstructed. I don't think that was a lie. You know, like, I think it was probably obstructed. But like, if that hadn't been the case, how would this cop have pulled over Olsen at DHS's behest? You know, like, you can pull somebody over for a license plate obstruction and then maybe, oh, wow, you just find all of this weed as well.

12:56But it's antithetical to how the justice system usually works, when it's actually the other way around. Well, I mean, I think the case, it's like maybe they would have pulled him over for going one mile an hour over the speed limit or something like that, or a rusted taillight or whatever. It's like, as we sort of have learned, it's like, it doesn't really matter what he was pulled over for. This reminds me of the story I did about Flock a few weeks ago related to Flock being used as a pretext to pull someone over for weed. Because in this case, the details are also somewhat sparse, but it's like this guy was pulled over for an obstructed license plate and he was charged with DUI.

13:44He was charged with intent to distribute, but it's not clear that that's the crime that DHS was actually looking for. I think that from what we can tell, it doesn't seem like he had kilograms of weed or anything like that. Like it was, he had some amount of weed. Weed is not legal in Wisconsin. And so it's like, it's, you know, it's unclear whether like this led to the type of crime that DHS was expecting. He doesn't seem to be, he's not a big time drug trafficker. I'll say that, you know, If that's what DHS is looking for, this Olsen guy doesn't appear to be that on the available evidence yet.

14:33But when I sent you the document, what did you think? I just immediately thought this is crazy. I think that we've written a lot about predictive policing in general, which is a term that is controversial because often it's tied to surveillance systems, to AI systems. And the thought of predictive policing is literally like minority report pre-crime vibes where it's like, well, the patterns of your behavior, we've detected that you are going to commit a crime and therefore we are going to intercede before this crime can be committed. Or I guess in this case, it's like our databases, our intelligence suggests that you may have committed a crime.

15:22we don't know exactly what that crime is but we are going to find a reason to pull you over and I mean it seems to me that like this is happening more often at the border although I'm not exactly sure um but yeah I mean I thought like this is crazy it's very dystopian it's sort of like uh there have been companies that have tried to do this before there's one called predpol which is like predictive policing literally um that's based on just like crime statistics in a neighborhood essentially um and i actually don't know this current status of this but this is the u.s government doing it the federal government working with local police it's just like a very insane thing um but yeah do you get a sense of like how widespread this program is so So there was some really, really important reporting somewhat recently from the Associated Press where two reporters there did a massive project where they, I mean, they revealed the Border Patrol was doing predictive policing in the first place.

16:27And that was very much focused on automatic license plate readers, like we've spoken about a million times. But these were a little bit different in that, if I remember correctly, it's a very long investigation that people should definitely go read. It's linked in our story. But in some cases, these were Border Patrol license plate readers hidden in traffic cones and traffic bollards and that sort of thing. And it created this network of license plate readers that Border Patrol were looking at to then predict, hmm, this vehicle is going to be suspicious based on where this person has gone and where they appear to be going and their past movements and all of that sort of thing.

17:13So that's an amazing investigation. They traveled all over the country to do that, definitely go read it. And now we're adding, oh, it's not just license plates. they're looking at financial data as well, which to be honest, to me, is like even wilder. Like, we kind of know that license plates are being used all the time. Again, Flock, Axon, Motorola, Focada, whatever. They're all over the place, right? But the idea that federal agencies are somehow just like, you know, panning for gold through financial data to then find somebody to be like, hmm, let's pull them over is crazy. And this isn't, I don't think this is in the document we got, but I imagine that this person, yes, looked at, sorry, this Border Patrol agent looked at financial data.

18:03Presumably, they also did look at license plate data to identify Olsen's vehicle in the first place. I think that's sort of just a given as well. Yeah, I mean, for this, this also raises some questions for me about federal efforts to police weed in places where it's legal, which we don't need to get so far into. But I think that a big part of the legal weed industry, which obviously federally still illegal, has been that they haven't been able to access bank banking stuff. and that had, I know that the weed industry has developed ways of using the banking system a little bit. And this raises questions for me of like, well, if they're using the banking system in any way, or if there's like tons of cash being withdrawn, things like that, like, I don't know, is this an effort by the feds to police weed in a different way?

19:01I think we don't have enough info about how this is working to say much more than that. But the fact that this guy was ultimately pulled over on weed charges is pretty interesting to me. And are we aware of anyone else that this has caught? Yeah, so... Well, I'll say first that there seems to be other PIT teams, PIT units, whatever we're calling them now, P-I-T-T's or whatever. But like this one focused on one of the north of the border. But then if you go through court records of a related case, which I'll get to in a minute because that actually answers your question. But you go through those, there's a bunch of text messages between law enforcement agents and Border Patrol agents.

19:50If you take one of those phone numbers and then you type them, if you take one of them and you put them into WhatsApp, the profile photo of that apparent border patrol agent is like a law enforcement patch, you know, where they have all of these symbols. Maybe it's a physical thing they put on their bag or their uniform or whatever. And it says tech ops PITT relating to a sector down south by the Mexico-US border. So already now, we've identified two PITTs, the one in the north related to the Olsen case, and then one down low thanks to this WhatsApp profile. I would say that Customs and Border Protection, which is sort of the one above Border Patrol, they have 20 sectors.

20:43And I did ask Customs and Border Protection, well, do you have like a pit in every single one of these sectors? And they didn't answer that. To go back to your question about who else this has impacted, This is mentioned extensively in the Associated Press report. It was also covered and obviously represented by the Institute for Justice, which is this sort of civil rights litigation organization, which has done a ton of stuff on automatic license plate readers. In that case, a cop pulled over somebody who was just like straight up innocent. They hadn't done anything and they ransacked their vehicle.

21:22They went through all of it. And then this target victim was sat in the passenger seat of a cop squad car talking to the officer who pulled him over. And he basically spilled the beans. Sort of like, yeah, this is based on intelligence and I'm looking for the big fish, blah, blah, blah. And that lawsuit, because this person who was unfairly pulled over sued, that opened up this whole new world of, oh, okay, there's something going on here. And that's how I got the phone number for this other PITT team. And I imagine how the Associated Press sort of learned of it in the first place, potentially.

22:05Maybe speaking out of turn there, I'm not actually sure how they figured it out. But yes, the point is, the case we spoke about and covered the Olson case, yes, he was charged with a DUI and then charged with a possession with intent to supply alleged crimes not proven in the court of law, not a perfect victim by his own admission when I've spoken to Olsen, you may think, well, the program is doing its job then because it's getting these people moving weed. Okay, well, what do you think about the innocent guy who was pulled over and didn't have anything at all? And just very briefly, I do think that there's too much emphasis sometimes in coverage or from the public of the need for a perfect victim.

22:48Like, I think that some people may only care about this story, or maybe some outlets would only cover this story. If it was a purely innocent person pulled over because of Border Patrol's predictive policing, I kind of, like, don't care. Like, it doesn't really matter. Sure, maybe this person was, maybe they are guilty of the DUI and the other charges, whatever. This was a vehicle for revealing more details about the predictive policing units, And I think ultimately that's what's important here. Yeah, agreed. Should we leave that there? And then after the break, Sam is going to ask you a bunch of questions about a new piece of AI fashion.

23:30We'll be right back after this.

23:37The difference between AI that's genuinely useful and AI that's just another thing you have to manage comes down to one word, context. If you're constantly explaining your company, your projects, what happened in the last meeting, and where everything lives, you're creating more work just to get AI to do that work. That's what I like about Notion. Notion is the AI workspace where your team's knowledge, projects, and agents all come together in one place. Because everything is connected, Notion has the context it needs to actually help your team do its best work. So instead of spending half your day searching through different tools trying to find an answer, Notion can help answer questions instantly, automate busy work, and keep project moving.

24:16And I think that's the version of AI people actually want. I use Notion to upload a lot of information about the articles I'm working on, a lot of 404 Media's background information, so that it's all in one place. That way, the other folks on our team can see what I'm working on, and I can keep track of what I'm working on as well. I've started using Notion over the last few months and I've noticed my productivity soar and it's been a lot easier just like keeping in touch with the rest of the team. This isn't some niche experiment. Notion has over 100 million users. It's trusted by 98 % of the Forbes Cloud 100 and 94 % of the top 50 AI companies, including OpenAI, Cursor, Lovable, and more.

24:54Learn more about how Notion can support your business at notion.com slash 404. That's all lowercase letters, notion.com slash 404 to try Notion today. And when you use our link, You're supporting our show. This message is sponsored by Raycon. I feel like earbuds usually make you choose between two things. Either you get to hear what you're listening to, or you get to hear what's happening around you. But there's lots of times when I actually want both. That's why I've been using Raycon's Essential Open Earbuds. They sit just outside your ear canal, so I can listen to music or a podcast while still hearing the world around me.

25:28If I'm out walking at the gym or just doing stuff around the house, I'm not completely disconnected from what's happening. This is super useful in Los Angeles where it's kind of dangerous to be a pedestrian or if I'm biking around or walking my dog, I can tell what's going on while still listening to my music. And they're really comfortable as well. They're ultra light and the flexible multi-angular ear hook adjust to your ear. So they stay put without feeling annoying after you've had them on for a while. They also have multi-point connectivity. So I can connect them to multiple devices and switch between them without constantly pairing and unpairing everything.

26:01Then there's the battery, eight hours of playtime and up to 36 hours total with the charging case. I've been using a few different types of Raycon headphones for different use cases. I've been having a good time with them. I wear them when I'm listening to podcasts, when I'm on phone calls, when I am listening to music, when I'm walking my dog, when I'm biking, when I'm working around the house. I really like them. Raycon has over 3 million customers, and there's also a 30-day money-back guarantee. 3 million people can't be wrong. The Essential Open Earbuds are the perfect companion for your fall reset.

26:35Go to buyraycon.com slash 404open to get 20 % off. buyraycon.com slash 404open for 20 % off. Thanks to Raycon for sponsoring. The 404 Media Podcast is brought to you by NordLayer Browser. NordLayer Browser is an enterprise browser that secures online work without getting in the way of productivity. It gives IT teams full visibility and control over every web-based AI tool, SaaS application, and download across the organization, while making sure employees can still enjoy a seamless browsing experience. Designed for today's distributed workforce, NordLayer Browser helps companies protect access to their internal resources and prevent accidental data exposure.

27:15Go to nordlayer.com slash browser slash 404podcast and use the coupon code 404podcast10 at checkout to get NordLayer Browser at 10 % off today. That's nordlayer.com slash browser slash 404podcast and use the coupon code 404podcast10. So for fun, I googled myself last week and found my home address, my phone number, and a relative I'm fairly sure I've never met all on directory sites I never signed up for. That's a data broker economy for you. They scrape your info, package you up like a trading card, and sell you off to whoever's paying, which is exactly how identity thieves go shopping. Bad luck when your entire job is not getting doxxed.

27:56So at 404 Media, we use Incogni. Instead of you spending hundreds of hours filling out opt-out forms across 40 browser tabs, Incogni gets to work by sending legal removal requests to over 420 harmful data broker sites. It's independently verified by Deloitte, resubmits requests every 30 to 60 days as brokers try to quietly relist you, and you can track every request right from the app. On the unlimited plan, the exposure scanner digs up the weirder stuff across the wider web, like old forums or random blogs or obscure directories. It surfaces all your exposed profile links right in your dashboard, where you simply select which links you want taken down, and Incogniz's team of human privacy experts will handle the manual corporate takedowns for you.

28:38Take back control today. Go to incogni.com slash 404media and use code 404media for 60 % off an annual plan. That's incogni.com slash 404media, promo code 404media, risk-free with a 30-day money-back guarantee. They can't harm you if they can't find you.

28:59All right, we are back. Sam, do you want to take this one, the Jason ring? Yes. So Jason, as the fashion experts on the team, let's get into fashion. Actually, I think, Joe, you're more fashionable than I am most of the time. I appreciate that. You guys are mogging me constantly. So yeah, this headline is, this digital camouflage shirt confuses AI-powered surveillance cameras. Yeah, so we've heard like adversarial fashion a bit in the past. It's something that's been around for many years, but you had the chance to kind of go and experience this particular project in person. So yeah, do you want to just kind of walk us through what happened when you went and checked out this?

29:51It's a shirt based on the headline, as we know. But yeah, what did you see? What happened when you like put it on and stood in front of a camera? Yeah. So essentially this shirt is, it's like made up of a bunch of blobs and we can talk about how it was made in a second, but it's like green and black and pink and red. It stands out. Like it stands out a lot. It's like Dan Flashes. That's the thing. I think this is why it resonated with people so much is because the pattern is so complicated. So yeah, if you don't know, please look up the, I think you should leave skit about Dan Flashes because it's exactly that.

Read the full transcript

30:31But basically it's designed to camouflage a person from an AI object recognition camera. So basically these are cameras that do object recognition, not facial recognition. So it identifies people as a person. And when you put this shirt on, it can no longer identify you as a person. It confuses that camera. We can talk about how it does that in a second. But basically, these cameras are used quite often in police surveillance and in private surveillance as well, where they can send an alert. If a person is recognized, they can also determine whether someone is laying down. So like if a homeless person is trying to sleep somewhere, they can also identify like different types of actions.

31:23So like fighting or anomalies or like someone loitering for a long time. And so this camera or this shirt is designed to confuse those types of cameras. And what this guy, Simon Weckert, did was he is running an open source object recognition algorithm called YOLO, which is very popular. It's like the starting point for a lot of different surveillance systems. Flock uses YOLO, for example, like a type of YOLO for its license plate recognition. And then a lot of other surveillance companies use this as a starting point. And so he, like in his studio, he is running a camera that's running this algorithm.

32:13And what he did was something called Gradient, Gradient something. but no basically what he does is he pits like an algorithm against this and so like if you're running the software you can tell the confidence score of what YOLO is seeing and so he will see like YOLO will say what like with 99 % confidence this is a person and then he designs a pattern and he runs that by YOLO and they say okay like 98 % sure this is a person and then he keeps altering the pattern, which is basically made up of random shapes until that confidence interval goes down and down and down and down and does it over and over and over again until it can no longer recognize you as a person.

33:06And he succeeded. This shirt does work for the YOLO open source algorithm. I was standing in his studio in front of the camera when I wasn't wearing the shirt. there was a big green box around me that said person. When I put the shirt in front of me, I was no longer recognized. And so, I mean, he's making these, he's selling them for, I think, like 80 bucks each. Like it is an art project. And we can talk more about like why he did this, what the purpose of art is in general, what the response has been, because this went extremely viral. Like not just our version, like our version on the website did okay.

33:49But I think on Instagram, like, I don't know, maybe a million people saw the video. But then I also saw like many other versions of this. Like people were talking about this far and wide, kind of like, what is the point of something like this? Is this going to give the police or surveillance companies incentive to improve their algorithm? So on and so forth. Is this Dan Flash's IRL? There was like many things that people were debating. I did see a surveillance company specifically use this in marketing where they were like, hey, look, our people detection algorithms are not fooled by the shirt.

34:29Like it was an actual like piece of marketing material on LinkedIn, which I know just felt a bit lame really to do that. Yeah, that is super lame. Good for them, I guess. but I feel like the thing like the like you said like this is an art project and a lot of the times with these kind of like so-called adversarial fashion projects I almost feel like and I think some of the artists who have done them have said that the point is not necessarily to be invisible walking around town for years years to come it's like the shirt's not going to work once the technology gets slightly better so it's not like this is an invisibility cloak forever it's more more like a conversation starter about the fact that you are being watched everywhere is, was that kind of Simon's take on it also?

35:20Like, was he, is he trying to make it more of a statement about surveillance than he is like, this is a functional item or does he think that this is like, everyone should be owning one of these shirts? Yeah. I mean, this is part of the article, but I didn't go that deep into it because I feel like people need to like understand this implicitly. It's like the point of this shirt is not to put this shirt on and then go commit a bunch of crimes. You are wearing a super bright shirt. Like it's like not, I think that that's part of the point is it's like super bright. It's you stick out like a sore thumb.

35:59It's like showing, like finding a person who commits a crime, who's wearing this shirt is not going to be like terribly difficult. the point also of like we've written about and seen shirts that have like license plates all over them like fake license plates and it's like to quote unquote poison the data or confuse a camera or whatever and it's like the point of those is not to go go poison flock forever like that doesn't really do anything what it does do is when you're out in public people will be like yo, why is your shirt so weird? What's going on here? And then you can say, oh, this is designed to confuse an AI-powered surveillance system.

36:44And then they'll go, whoa, what is an AI-powered surveillance system? And then you can explain what it is and raise awareness about that thing. And that is the entire point of this shirt of a lot of adversarial fashion. I mean, I think that there are some that are a bit more practical, like wearing a full on mask at a protest or something that where you think there's gonna be facial recognition or something like that. Like, sure, that has kind of a more practical use. But something like this, where it is an art project, where it specifically probably doesn't work against higher tech, like higher tech implementations of AI cameras, like that's part of the point.

37:28One of the points is, we don't know specifically which algorithms are running on these cameras. That information is not released to the public. But this is kind of like the best that we can do is like using these open source systems. You can attack that and you can sort of show how powerful these open source systems are and then start a conversation about the fact that the ones that the police actually have or that companies actually have are better, more powerful than this. And then we can be like, well, maybe this is something we want in our society. Maybe it's not. Like that is the entire point of this.

38:08That's the point of art. It's the point like to provoke a conversation, a response, et cetera. And this has been like wildly successful in doing so. But I think that there are a lot of people, like a lot of people understand that, but then there's been a lot of response that's like, oh, you're telling the surveillance companies that this works. And then therefore they're just going to update their algorithm. And then Simon is going to have to update his shirt. And then like, this is not going to work for that long. And it's like, that is the entire point. It's like, it's to provoke a response from the surveillance companies.

38:47It's a pro response from the people wearing it. It's a pro-correspondence from society. We cover this type of thing as journalists to raise awareness about the fact that it's happening so that people can have a conversation about whether this is something that should exist or not. It's like we all have our role to play, basically. And it's like the role of this type of art is to be like, yo, like here's, If you're not interested in reading thousands of words about how this sort of thing happens, maybe you're interested in watching a 30-second Instagram video that clearly shows how this technology works.

39:26Or maybe you're into fashion and you're like, whoa, this is the ugliest shirt I've ever seen. Why does this exist? And you can have a conversation about it then. And it's like, the fact that some people can't understand that is like mind-boggling to me, but I guess many people are stupid. Yeah, it's not meant to be taken so, so, so. Literally, if you want to be safer and more disguised against facial recognition in public, just wear a hat, like a baseball hat over your head and dark glasses and a mask. That's the easiest way. But even then, a lot of these cameras can detect they can identify you through your walking style and gait and lots of other things.

40:07So, I don't know. I feel like the commentary of it all is still important to have. I hesitate to say it's ugly. It's kind of cool looking, I think. I would buy one. It's not ugly. I would wear it. It's definitely different. I think it's cool, but a lot of people are like, this is a really ugly shirt. That is what some people have said. But it's just like, they just don't appreciate complicated patterns. I think the people saying that know subconsciously or consciously that they just can't pull the shirt off. Yeah. Yeah. I think we should all get one, probably. I also saw somebody in the comments of this story on our site say, this might be a great way to get hit by a self-driving car, which is like, that is like, I think we should leave skit.

40:53Like, I was avoiding getting facial recognition and I got hit by Waymo. I don't know if that's how it works, but I'm sure that's not actually how it works. But I think that's... LiDAR. I mean, it's a different technology, but it is like similar, as in they're detecting objects and things like that. So I would imagine whatever Waymo is running is more sophisticated than this off the shelf, whatever. And that's the other thing. It's like, oh, by showing that these things can be attacked, the surveillance companies are going to try to improve their algorithms. And it's like, they're trying to do that anyway.

41:29They're improving their algorithms all the time. Like that is the whole, yes, that's what they do. Yeah, totally. Cool. Well, okay. We're all going to order a shirt and wear it on the next pod and confuse the recognition systems of YouTube, I guess. Well, yeah, and Riverside, the AI. It'll break. It'll completely break. In the podcast platform we use. All right. Yes, as Sam said, we'll leave that there. And if you are listening to the free version of the podcast, I'll now play us out. But if you are a paying 404 media subscriber, we're going to talk about our third year anniversary. We had the panel, we had the party.

42:08Both were amazing. There will be a recording of that to come at some point, but we'll discuss the details of that after the break. You can subscribe and gain access to that content at 404media.co. As a reminder, 404 Media is journalist-founded and supported by subscribers. If you do wish to subscribe to 404 Media, and directly support our work, please go to 404media.co. You'll get unlimited access to our articles and an ad-free version of this podcast. You'll also get to listen to the subscribers only section where we talk about a bonus story each week. Another way to support us is by leaving a five-star rating and review for the podcast.

42:50That stuff really, really does help us out or just tell a friend. This has been 404 Media. We'll see you again next week.

43:02Let's go. Thank you.

From the publisher

Take control of your data footprint risk-free with a 30-day money-back guarantee. Go to https://incogni.com/404Media and use code 404MEDIA for 60% off an annual plan. That's code 404MEDIA at https://incogni.com/404Media

We start this week with Joseph’s story about the secretive predictive policing units inside DHS that are analyzing Americans’ financial data and having local cops pull them over. After the break, Jason tells us about the funky shirt designed to fool AI-camera systems. In the subscribers-only section, we talk all about our third anniversary events. 

YouTube version: https://youtu.be/o0AG6A5MIRg

RSVP: Killer Technology With 404 Media

A Secretive DHS ‘Predictive Policing’ Unit is Analyzing Americans’ Financial Habits and Pulling Them Over

This 'Digital Camouflage' Shirt Confuses AI-Powered Surveillance Cameras

We Are Going to Be Okay: A Three Year Anniversary Events Recap

Learn more about your ad choices. Visit megaphone.fm/adchoices

More from The 404 Media Podcast

All 164 episodes
DHS’ Secretive ‘Predictive Policing’ Unit Pulling People OverThe 404 Media Podcast · 43 min
Listen in VO