Skip to content
PodSubs
ShowsPeopleGuestsJobsNewsEvents
More
BlogPodCred
Menu
ShowsPeopleGuestsJobsNewsEventsBlogPodCredNetworksSign in
Join nowSign in
  • Shows
  • People
  • Jobs
  • News
  • More
    • Guests
    • Events
    • Blog
    • PodCred
    • Networks
    Join nowSign in

The Low Down

The Low Down artwork

LG Caught Spying in Your Living Room, Miami Crypto King Gets Life, and the AI Apocalypse Debate

Sep 18, 20261 hr 17 minAudio
ShareXLinkedInFacebookWhatsAppEmail

Welcome to The Low Down, the best show on the internet for hackers The Low Down is presented by Maze. LinkedIn: https://www.linkedin.com/company/mazehq/ X: https://twitter.com/Maze_Security Follow Us! https://www.instagram.com/lowdown.pod This week we're diving into the surveillance nightmare coming from your living room, catastrophic phishing campaigns targeting crypto users, and Microsoft's record breaking patch cycle that defenders need to know about. Today we're talking about: LG's Privacy Nightmare: Your TV is Spying on Everything Gamers Nexus drops a two hour expose revealing LG TVs are actively recording and transcribing all room audio even when listening mode is disabled and the device is powered off. Breaking down how they got root shell access and found plain text transcriptions of private conversations being cached on the device with four gigs of memory ready to phone home. Why the microphone continues recording even when unplugged from the network and what ACR automatic content recognition really means for your privacy. The Economics of Smart TV Surveillance LG executives caught on stage at advertising conferences bragging about owning the glass and extending the advertising experience across all devices on your home network. How they're doing network scans to enumerate every Windows PC, phone, and IoT device in your household and correlating that with what you're watching. Why the profit margins on TV advertising are in the 60s with SaaS level multiples while TV hardware prices have bottomed out over the last 20 years. The Remote Code Execution Problem Nobody's Talking About Beyond the privacy concerns, LG TVs have multiple unpatched remote code execution vulnerabilities that turn your living room microphone into a potential threat actor playground. Why teachers are panicking about 85 of these devices deployed in classrooms and the student privacy implications. How Chinese threat actors have proven they'll exploit residential devices for proxy networks and credential theft. Trezor's Catastrophic Phishing Campaign Trezor customers hit with one of the nastiest phishing campaigns of the year where the call was literally coming from inside the house. Breaking down how their official email provider was breached with valid TLS certificates and domain signing making the phishing email appear completely legitimate. Why this was perfectly timed after the cold card STM 32 entropy vulnerability had everyone primed to believe their hardware wallets were compromised.

Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.

Transcript

From the show. Tap a time to play from there.

Good morning, hackers. Welcome back to The Lowdown presented by Maze, the best podcast on the Internet for hackers. Matt, how are we doing, man? Listen, before you start, I want to go, okay? Your boy is a little tired. Why are you tired, little level? Great question. I have two kids. One is three. One is one. And we decided it would be a good idea to go on a beach vacation with my two children. What do you think? What are your thoughts on that? Man, I was just, I was watching our podcast back from last week and I, I was the rookie dad complaining about going to a hot soccer game without an umbrella.

Now you're the rookie dad going to a sandy beach, sunny location with two kids that don't want to wear suntan lotion. Don't want to wear the rash guard. one or both are still taking naps. Definitely one. One for sure. Both should be, but he refuses. So right, right, right. And so what you're doing the like, leave a baby monitor in the hotel room and like sit somewhere or you're in the hotel room. Oh, it's actually rad for me because I'm pasty white and antisocial.

So like when my kid has to go down for a nap, I'm in there and I'm like, oh no, I got to watch the baby guys and they leave me alone. It's fucking awesome. Okay, okay. That's like a good excuse for you. I got it. I don't know. We've done this. So I was telling you before we recorded the trick is, and you're almost at that age is to travel with their friends. Yeah. And then like, it just starts to get a little easier because you're, then you're not like the only source of all of their attention and energy is pointed at you and your wife. Right. Right. And it's like, Oh no, look, another, you know, three, four year old, like go, right.

Go outside for 16 hours, please. Goodbye. Yeah. Yeah. Like four is probably the good age for that. Like, it's worth this weight in gold. Find another couple that you really vibe with that is a kid the same age. And it's just like, ooh, okay, things start to magically happen. And then vacations become vacations again. Slightly more, slightly more. We'll get there. Someday. Someday I will have my vacations back. Yeah. Yeah, I acknowledge very lucky, very privileged even be able to go on vacation. It's just yeah. Yeah.

As one of my friends told me recently, it is literally just parenting in a hostile environment. You know what I mean? That's a really good way to put it. It's really good. I know you're probably wondering, you know, guys, isn't this a hacking podcast? We're talking about kids. Well, here's a problem, right? You start as a hacker and eventually you have a family. And then you have to figure out how do I mentally balance being a hacker and a parent? And then these worlds kind of weave together a little bit. So I was thinking about, you know, exploitation and zero days on the beach while my three-year-old cried about his Mickey Mouse falling into the pool. So it's an interesting dichotomy with it.

The way I like to phrase it, I mean, we're not just hackers. We're also like entrepreneurs, I guess, right? We both run our own businesses. Yeah, I know. We're not like we're not VC backed. you know, software companies, but we both are self-employed. Yet, yet, yet. Ooh, little foreshadowing. No, but I like to say that we're doing startup life on hard mode with, you know, with little kids at home. It's like when you're 20 and you can survive off of like Pringles and ramen and caffeine and like cigarettes and like no sleep.

It's like, oh yeah, of course you could do a startup. Like when you're in your like early for your case, late for my case, 30s with like small children and, you know, mortgage and all this kind of shit. It's like, oh, wow. Yeah, no, I got to like go to bed at a reasonable hour, which I did not do last night. And so I'm also kind of feeling feeling it with you. You know who else is feeling it right now? LG? The sponsor of today's video. Just kidding, not yet. Yes, LG, bro. Listen, guys, we've been talking a lot about certified pre-owned, right?

Certified pre-owned. And the topic of that conversation was a lot of these TVs are, or not even TVs, a lot of these devices you buy online, routers, Amazon Firesticks, cheap Chinese knockoffs of Amazon Firesticks, are showing up pre-owned. The issue there, or kind of like the solution to that problem, is that these are brands you've never heard of. There's a fundamental trust misalignment. You don't trust these brands. You, therefore, should be able to trust other brands. But the issue here is we're starting to see a shift, fundamentally, a paradigm change in how brands are behaving.

And LG, I think one of those brands that we thought we could trust for a long time, has kind of turned the tide. Matt, I see you tensing up. What's going on, man? I am so angry about this one. I have so many big feelings about this one. I went, I, like I live streamed yesterday for the first time in a few weeks of all the travel and everything. Right. I was raging, right? Raging because exactly that, exactly that. These are $3,000 devices. Imagine you fired up your MacBook Pro that you just spent way too much effing money on or your Mac Studio that took 18 weeks to get to you because everyone had to install OpenClaw a few weeks ago and it cost you, you know, thousands of dollars for this premium piece of hardware.

And you opened it up and it was like... Please install McAfee and we're going to spy on everything you effing do. And we're siphoning all your data off so that we can better advertise to you. No, this is why Mac is Mac. And like the $400 Windows laptop at Best Buy is the $400 Windows laptop at Best Buy. Right. What's the difference? Well, it's not just the fact that these things work and like there's speed and like hardware specs. No, the equivalent hardware spec Mac. to like Windows device is not the same price.

The reason that you're paying this luxury tax is for this luxury experience. And what do we get out of LG right now? Well, to kind of butt off of that, there's also like the psychological element of like premium pricing, right? Like if I'm gonna spend $3,200 on a thing, there's a marketing choice that they do there to try to make, like trick me into believing that I'm getting a good product. And I think with Mac, you really do get that. I think you're paying for high quality hardware, high quality software, good security, whatever. And so for certain other brands, you would assume that that translates, right?

Oh, I'm getting a $3,200 TV. Okay. Surely that means I have smooth feel on the remote. All the apps are responsive and, Oh, I'm paying a high price. So therefore it doesn't need to be subsidized by advertiser ID. Right. And the, Lo and behold, not the case. Yeah, I guess if you've missed what we're talking about, which I don't know, I've made content about this before the podcast and the video that we're going off of is two hours and 15 minutes long with like last I checked over three million views, I believe.

I mean, this is obviously pushed a hot button for people. The Gamers Nexus crew, which has taken a turn in terms of YouTube direction in a way that I am very supportive of. They've got hackers and stuff in there. They were walking around a hardware lab. They were doing stuff that looked like you would have loved, right? They got the backs pulled off of these consumer electronics, and they're doing very low-level things. Name of the channel. In the channel, by the way. And so, yeah, like kudos.

I watched this. I watched all two hours plus of it more than once. I've like combed through it because I've made a bunch of content about this. I've written about it. I've tried to like summarize each of the specific findings that I thought were super spicy. So let's set the stage, right? I'm angry. Obviously, the backdrop is LG's kind of expose about some privacy concerns. There's also security concerns, by the way. We can talk about some remote code execution in your TV that we don't have the details of yet. It's unpatched. That's great.

O-days. So if you don't listen to the rest of the podcast, just go take... If you have an LG TV, just don't put it on the network. How about that? Right? That's like the TLDR. The video goes into a few things. We could also critique some of the video if we wanted to. Did you watch it? Have you gotten to catch up on it? Let's just run down what we're actually talking about. The TLDR, before we go into details here, is that LG TVs have been caught spying actively in your home. And I know that's kind of been like... A thing that's assumed, unfortunately, like, oh, TV has microphone, TV has ads, therefore it's listening to me and it's going to like sort that data up and push it somewhere.

Right. And again, it's an unfortunate world that we live in. We're like, that is now the assumed posture of our technology. But there really has been no hard definitive proof in a lot of these cases. Well, so Gamer Nexus went in with I don't know the name of the other researchers, but they did, again, a two hour expose. showing by exploiting a vulnerability to get them a root shell on the TV, even when listening mode is quote-unquote disabled, the microphone is still recording and transcribing all room audio into plain text on the device,

and caching it for some reason. And then they also found that there are multiple remote code execution vulnerabilities, emphasis on remote, like over-the-internet vulnerabilities, in these TVs, right? The worst part is that LG was put on the spot and tried to address it, and they basically said, uh-uh, that's not true, that's not real, even though, like, literally they had a root shell on the device, and it was actively occurring. Gosh, it is so much worse than even summing it up could possibly make.

This is why it took two hours to go through it. So we're going to do our best. But that's the crux of it. The most egregious findings is that, A, did I even realize that my LG TV had a microphone? Yeah. No. Did I realize it had a wake word? No. Like if I say Siri, it's by mistake. Right. I never want a wake word for anything. Right. Like, you know, I've pulled all Alexis out of my house again. Small children at home make that a nightmare. You know, and yeah, I just don't like Internet connected microphones just like around all the time that I don't have like a decent idea of what's going on.

I think there's a lot of conspiracy theories around your phones listening to you and then advertising stuff shows up later. A lot of that has been at least not proven, maybe not disproven, whatever. But we haven't found a ton of evidence that that's actually happening, even though everyone kind of says it. We've all had that experience, right? Where you're scrolling Instagram and you're like, I was just talking about this yesterday. Like, what the hell? But this is actually truly happening on your TV, and they have evidence of it, of conversations that you're having going over the wire in clear text, including sensitive information.

There's absolutely no filtering, including when the device is unplugged from the network and including when the device is powered off, screen off, screen black. It is still recording audio. It has four gigs of built-in memory to shovel whatever it's recording away for the next time that it is woken up and put on a network. Kind of ridiculous. The acronym, you know, du jour on this story is ACR, Automatic Content Recognition. And the thing that like really pisses me off because they have a response to the Gamers Nexus thing, what we can get to and we can show.

The thing that grinds my gears is in the Gamers Nexus video, the executives from LG are like on panels and on podcasts and on at Cannes Lions and all these different advertising conferences bragging about the capability that they are now trying to like walk back after everyone's like yelling at them. They're like, well, ACR, what are you talking about? That's like, well, a that's opt in and they try to, you know, kind of hide behind that. And it's like, well, Okay, I have to breathe.

The opt-in, did you see the opt-in part? It's like the select all is auto-highlighted. They really shadily try to get you to opt-in to all of this stuff. It's dark pattern shit. It's the same argument people make events like every time Windows 11 pushes a new update, you have to re-go through the freaking obstacle course that is getting your device started automatically. And every time they change the default settings of Cortana microphone access, like, Oh, is co-pilot on? Are you sure you don't want it on? Like all of that shit. It's the same dark pattern shit. Yep. Yep.

And like no one is consciously opting in to like advertise to me more daddy. Right? Like that's not like no one is reading the 50,000 page doc and going, yes, like, please may I have another, like take all of my living room conversations and serve me a, like nanosecond speed AI contextual ads of it. And that's what he's literally on stage bragging about. The president of LG ads who LG ads is a private separate entity from the public LG.

And that's important because. This guy is barred from the SEC from running a public company for doing scams in the past. Wow, that I was not aware of. That's insane. Yeah. So this guy, the president of LG ads is barred from the SEC from running a public company. So he goes on. If you watch any part of the video, I think this is probably the most revealing of what's going on. It's like early in the video. the exact there's like a clip of a bunch of these different LG ads, again, separate entity executives talking about their capability. And they like, they like to use this phrase.

We own the glass. Yeah. We own the glass, which just gives me the creeps. It's like, we own the glass. Well, I thought I just paid $3,400 for the glass, but okay, cool. Do you? You bought a license for the glass, good sir. Right, right, right, right. And cue GTA 6 hacker issue. What's going on? We don't own anything anymore. But they're literally bragging. They're like, people sit and watch TV, and they sit on their phone. And they then know the other thing that I don't think you mentioned in your summary is that they're also scanning the entire local network for all the other devices on the network.

So not just audio capture, but they're also doing basically NMAP scans because it's not even just like, oh, this is the device. They even had examples of like, it's a Windows device with these services exposed. Yeah. I wanted to ask, so I didn't go into the article or watch the video yet, I read kind of the headlines on the Twitter post about this. When they said it's doing network scanning, are we talking about actual NMAP TCP SYN scanning? Is it doing SSDP? Because when people say network scanning, that word means so many things.

There are certain protocols that are there by design to enable scans, like SSDP for example. simple service discovery protocol. It's a protocol that exists literally to enumerate a network on purpose. It's not nefarious action. But if they're doing, like, TCP, SYN level, and map-style scanning, that feels a lot more nefarious. I'm curious, not that it makes it any better, it just kind of, I think, helps paint LG into the proper hacker corner that they're acting like, right?

Yeah. And they, they might've gotten into that in the video and I just haven't committed that specific detail to memory. Um, but I do have a screenshot that I can pull up. Um, I could show you, let me just like post it in the show notes. We can put it on the screen. Um, that, uh, I put, I like wrote, uh, an X article I've never written. I don't write like Twitter articles or anything, but I wrote one summing up this video. Cause I was like, Hey, not everyone's got two hours to watch this. Like, here's all the findings that I think are interesting. And one of the screenshots from it is all the devices on the network.

Right. And so I don't I don't like have this specific like, well, what on the Wireshark like did they see actually happen? Because they did do that right in the lab and all that. But the the list here, they redact a bunch of it. And the one that I have, like I left highlighted. kind of told me a little bit more about the scan that went on. So it says Vitaly's computer is the name of the computer, Windows PC, parentheses, team viewer running detected via MDNS.

So it's like not just, oh, there's this Windows box on this network at that IP, but it has services running on it. Right. So that seems a little bit. lower down. And then, yeah, I just, they pulled this info from the TV logs and the TV phoning home, like network traffic. Right. So it is phoning home with like every other device on your network. They've like walked a lot of this back and said like, well, this is just like the access, the permissions that we need to be a smart TV.

Like we need, probably your phone could be a remote, right? So in order for your phone to be a remote, it needs to be able to like, understand that your phone is on the same wifi. And like, there's like a pairing handshake thing that winds up happening, you know? And like, that's probably what they're talking about, right? But then you don't need to like send the name of all these devices home, first of all. Second of all, back to the execs on all the podcasts bragging about everything. They're saying we understand like on, like on purpose, they're like, we know every device in the LG household.

We can extend the advertising experience off the TV onto your other devices. These are the things that they're flexing that they can do. So, like, no, you're sorry. We're just doing the things that smart TVs need to do. It doesn't hold water when you then go on stage and brag about being able to extend the advertising experience to the other devices on the network that aren't your LG TV. Yeah. As a person who invests in the stock market a little bit, it makes me very concerned that, like, companies are just out of ideas.

And, like, all we have left to do is milk the advertising market a little harder. And my question becomes, like, what happens when that... either kind of starts to get more found out like this, starts to get more heavily regulated, and all we have left to do is shove AI in things. Are we as a society out of things to build? Are we done? You've had the tinfoil hat animation show up on your head. Mine is we do not have a USA GDPR equivalent because too much of our economy and stock market depends on...

Exactly this advertising is propping up so much of the market that they fear our government fears. If we put too much consumer privacy regulation in place that the popsicle sticks holding this whole thing up start to chip away. Right. The Jenga block on the side of the whole thing gets kicked out. Yeah, it's not great. I mean, yeah, but look at like, have there really been any significant technological improvements in the last like 15 years? Obviously, RAM's gotten better, CPU's gotten faster, but it's all just been like the experiences have been the same and it's just been more data getting slurped up.

I mean, it's why I mean, AI is the answer to that, right? And that's why there's so much focus on AI specifically. Yeah. Hey hackers, before we keep going, I wanna take a quick break and say a quick thank you to our sponsors. Today's Lowdown episode is sponsored by our friends over at Maze. Maze has spent the last couple of years using AI to find and remediate vulnerabilities in the cloud. And now they've released Maze Code, AI agents that find and deeply investigate vulnerabilities in your code. Now we all know the deal. Software composition analysis tools and SaaS toolings are all historically really bad at just making more noise.

They find issues that may not actually be real and just create a lot of problems for your engineers to deal with that don't solve real security vulnerabilities. I get it. We've been there. Maze code is a really cool tool because Maze code actually knows how your code works. MazeCode investigates every finding with the context of how your code is actually deployed. Sure, you're using a version of libxml2 that's vulnerable, but the actual code path to hit that vulnerability is not exposed in your Docker image. MazeCode can find that and won't show you that that vulnerability is a big deal.

Every vulnerability that MazeCode reports comes with evidence that it's real and a way to fix it. Check out Maze at go.lowdownpod.com slash maze. That's go.lowdownpod.com slash maze. Thanks again, Mays. Let's keep going. Oh, that was another thing. Okay, hold on. I got to tell you this one. This one really pissed me off. The president of LG ads, again, talking at like some advertising conference, bragging about their capabilities, starts to talk about this ACR, automatic content recognition, which is recognizing the stuff that's on your screen.

And it's also recognizing the stuff you're saying around your screen, right? But it's also what's on your screen, which like... I don't know. I don't really think, oh, my TV manufacturer is going to know what I watch, even if I'm not using their app or something like that, just because they can do this content recognition of what's on the screen. Maybe people don't realize that, right? And that's not very transparent. He is out there saying... Their kind of milk, you know, their big technological idea is the exact scenario he described.

You're watching a bar scene and a certain brand of drink is like in the scene and there's a product placement in the show that you're watching. How does Bo Burnham say it? It's with the label facing out. Right. Right. Right. Right. And so that kind of scene is happening. And he's saying with ACR and LG tied to their AI platform for contextual analysis, capturing that content, they can then in real time serve you an ad for that particular drink that you just watched the actor drink in your favorite show on your phone.

Who is opting into this? You know what I mean? Like, no, no, no, no, no, no, no, no, no, no, no, no, no. Right? Wow. So hold on. So even if I use, like, an Apple TV and I push my Mac screen to my Apple TV to, like, work or whatever, they are still capturing the HDMI output of the Apple TV into their TV and selling me whatever. This is like what the ACR, this is like all the ACR stuff, right? If you use a lot of their apps, they even have more because then they've got like different DNS queries and stuff that they set. So like if you know how like all the smart TVs have like that fake cable shit that's like built in now that like feels like channels, right?

And it's just like NCIS reruns and like, you know, whatever. They showed the DNS queries like reporting back exactly what channel you're watching. I was like, that's not that big a deal, whatever. Okay, cool. I'm watching NBC or whatever. Okay. But again, just dark patterns of what you're saying. Then there's remote code execution vulnerabilities. So everything that we just talked about is LG being kind of annoying, advertising, data, brokery, weird behaviors.

But all the capabilities that we just mentioned that this hardware can do that everyone's putting in their living room and their boardrooms. I had teachers DMing me about this video saying we have 85 of these things in our classrooms. Really worried of how to approach this with my school board and all this kind of stuff. I'm like, geez, I don't envy you having to figure out. what kind of student privacy concerns you now are facing, right? Not like ignore LG and advertising completely.

The fact that this thing has remote code execution vulnerabilities and the four gigs of memory and of storage and microphones that are connected to the internet mean that a threat actor could do all of this very, very, very easily. And a bunch of the Chinese threat actors have proven that they have absolutely no problem hacking residential devices that have RCE on them for residential proxy stuff or for whatever else is going on.

So that's another major concern here. Yeah. Again, the whole we talk about this a lot, right? With these technologies, it's a question of acknowledging your threat model, right? Acknowledging what risk you're willing to take on as a person and building your life kind of around that. I think that no one should have to take on the threat model of like, oh, LG is watching me. Right. LG is spying on my family. And so in my opinion, I think what everyone should do, and again, not sponsored, but you know, hit me up, is buy like a 55, 65 inch, simple $800 Samsung TV with no microphone, with no camera, and strap in a fucking Apple TV, man.

And those things are hands down the best device that I've ever used. They're extremely responsive. And Apple, you know, there could be some breaking expose in five years on Apple as well, but they kind of posture themselves as like the privacy company. There's no real public indication that they're doing the similar kind of advertising, slurp up all your data, food that other companies are. So that is my home recommendation for TVs because it just feels like every company is getting caught with their pants down doing this kind of stuff. I have a feeling that LG is not going to be the last brand decimated in this way.

I used to be a Samsung loyal. Their whole operating system and ecosystem started to really just feel like crap for me, even though I'm also Apple TV pilled and like you, it's worth the money. It's responsive. It works. It doesn't advertise to you. It's great. But I don't know. I can't stand here and be like, yeah, Samsung. And then you're good because like their OS feels like crap too. Right. So like I can't imagine that they're not also doing stuff. The real like the end of the day kind of why this is all happening.

The economics of this is insane. Right. Like how many TVs do you buy? a year, right? 0.1, right? I think every, like every five years, maybe we have maybe, maybe. Right. I think I had, I think I had my other one at my, uh, at my old house, like for a decade. Right. And I was like, ah, it's good. You know, like they've kind of like, there was a while where they were iterating rapidly with like new features. And now it's kind of like, well, how thin do you need the damn thing? Right. Right. Right. Right. Um, And so that's part of the problem. And so the profit margins on these ad businesses are insane in the 60s.

And so they're like, you know, they're basically seeing SaaS multiples on like the ad business for a commodity hardware business where you're competing against like $200 Vizios or whatever the heck, you know. Best Buy store brand stuff is, is what you're competing against. And so have you seen that chart where inflation and everything is like... Cost of everything has gone nuts over the last 20 years. The TVs have just gone to... The TVs are at the bottom. They've bottomed out, right? And it's like, well, here's why, right?

They've figured out this... Well, yeah, we'll sell you a cheap device and then do this other stuff to you. But that's why it makes me so mad. The LGs are not... the cheap timu ones are not the cheap best buy one right it's like these are a premium device that you shouldn't be figuring out how to you know get me to install the mcafee subscription on the cheap laptop is like what it feels like here it's so annoying dude are you a cryptocurrency enjoyer are you a crypto person zero percent zero zero percent wow interesting

0%. Yeah, so I don't have a whole Bitcoin, but I have parts of a Bitcoin. And so I very lightly follow the cryptocurrency community, not so much like the meme coin, the shit coin community, but like the hardware community, right? I'm a hardware guy. I'm a hardware vulnerability researcher. I like exploits. And so... This Trezor email popped off, and I was kind of concerned that potentially my wallet, not a Trezor, it's an offset concern, I would never tell you what I have, may be compromised. But it turns out the whole thing was a phishing campaign. Matt, what are we talking about?

What's going on? Dude, this company is cooked. Like, I can't even understand how many different ways they're screwed, right? It's bad. Like, A... like crypto is a target for hackers for exactly these kinds of reasons, right? Once stolen, like is gone is generally people's experience with stolen crypto, right? It's like why North Korean threat actors are focused on it. Cause like, they're really good at targeting it, mixing it and all this kind of stuff.

We have like a number of crypto stories to talk about today, including like a massive crypto scrammer going to jail. But this treasure one, They, their customers just got hit with probably one of the nastiest phishing campaigns I've seen this year. Because the phishing, the call was coming from inside the house, low level. The phishing email came from their official, like, company email signed, like, TLS cert. Like, everything was good about this.

I'm seeing people dump the headers. I'm seeing people dump... They're like email security tools output on this treasure phishing email. And then what is treasure tweet? What is treasure tweet? Sorry. Oh, we're so sorry. Our third party email provider has been breached. Please be aware the email named quote, critical security alert, STM 32 entropy vulnerability is not coming from us and is a phishing attempt. Do not click any link.

And then in that thread, you could see, I mean, it's a pretty good phishing email too. But then also in that thread, this email, this, this company continues to be cooked because people then are asking the company Twitter account, which I hope that is an official communications channel that is still trustworthy because this is crazy. And someone's tweeting them with a picture of physical mail that got sent to their house with QR codes on it that talk about like their treasure wallet. And they're like, I don't know. How about this? Is this official? I don't know how this even got my current address.

I've moved a bunch of times, but here it is. And yes, I also got the email today and treasure replies, Hey, unfortunately, another completely different third party we use that for shipping also experienced a breach. And they were targeting our customers to get their addresses to send these kinds of scam physical mail to people. So Trezor has like a had a lot of security concerns, but B is doing a really piss poor job picking third party providers for very mission critical things, apparently.

Yeah, the reason why this is such a devastating blow is obviously having your TLS cert for your email compromised is like one thing, because that is supposed to be like the root of trust. Like mathematically, it is proven that it is me. Like that is the bar for trust. Okay, so that got compromised. Also, physical security, right? Physical email provider got compromised. Shit. Okay, so that's bad. But if you top all that off... The reason why this is so bad is everyone was already primed for there to be an STM-32 vulnerability in Trezor.

So going back a couple months ago, I'm going to look at the exact date here. This was literally last month, August. Cold Card, so this is another hardware wallet, was also compromised, quote-unquote, there was an entropy vulnerability. So again, I'm not a crypto expert. I'm just going to talk a little bit about how I understand the blockchain to work, okay? So for Bitcoin, for example, right? If you want to purchase or sell a Bitcoin, the way that it works is you have two keys, right? You have just like any kind of asymmetric scheme, any kind of PKI, you have a private key and a public key. Now, what you do is you have the ledger of how much money you have and you sign with your private key a transaction that says, I want to give this many Bitcoin to this other wallet.

That transaction. transaction goes out into the blockchain and other nodes verify it's a valid transaction. And you use that to keep the public ledger of who has how much money. The crux of that is that the private key obviously remains private, right? The private key being a fundamental piece of cryptography that is generated on a hardware wallet when it is instantiated. Okay. Now the is a function of this thing called a seed phrase. Okay, like any kind of crypto algorithm, the randomness is super important because being able to predict the seed phrase on this hardware wallet allows you to therefore predict the private key.

From a 12 or 24 word seed phrase that every hardware wallet has, you can therefore predict the private key. This cold card had this issue where they pushed a firmware update in 2021 that changed how the seed phrase was derived. It used a form of RNG that used an internal timer inside of the wallet at boot. which could relatively be predicted, right? The boot timer that it takes to boot up your wallet is probably going to have a fixed timer amount by the time you get to the RNG creations phrase.

And so because of that, it reduced the total entropy of the seed phrase generation step from 128 bits, which has, I think, what is it? A quadrillion, a fake made-up number, a very large number that could never be brute-forced, down to 40 bits of entropy, okay? Okay. Because of this, the way that a blockchain works is you are able to now guess, effectively, the seed phrase, therefore the private key, on the network of who – like, you could basically brute force these private keys, put them out to the network, and drain wallets for free.

Okay, this is a big vulnerability that happened last month. And so everyone's like, oh, my God, like very nervous about Trezor potentially being vulnerable to this. Ultimately, it was not. It was just a big, you know, phishing campaign. But it's just the worst case scenario, right? Real bug in a different manufacturer exists. This bug comes out, not a real bug, looks like a real bug. Oh, but it's signed by Trezor. I'm going to go act. And in doing that, you do a Bitcoin migration from your wallet. And oops, who you migrated to is not actually a real wallet. It's a fake account.

So if you are vulnerable or if you've been compromised with this, I'm very sorry. That is wildly unfortunate. Like Matt alluded to previously, the problem with crypto is there are no regulatory bodies. There's no FDIC insurance. So you're kind of just shit out of luck, unfortunately. Yeah, I think the space is like primed for scams like this too, right? Because it's, A, it's hard to tell what's real for most people. I mean, good fishing is good fishing. It works for a reason. Now you got people in like heightened sensitivity because there's like money involved and all this kind of stuff.

These kinds of things can happen. Case in point, the next two stories we're going to talk about. So number one, it seems like we have some people calling themselves white hat hackers who some other people might debate the color of said hat who have found a vulnerability in liquid network, which can you do you know much about? I actually don't. I have no idea what that is. Okay, it's a Bitcoin market. It's as much as I'm willing to talk about it. It's not a crypto guy either. But they were drained of a couple hundred million dollars.

And so the thing with crypto and the blockchain is there's a lot of people watching Bitcoin. a lot of these blockchains and can report what they're seeing in terms of like big anomalous movements before we ever actually get a news story about it. This is one of those such cases, right? So we have like some Twitter accounts. Liquid network was drained of approximately 4,000 Bitcoin, which is like $320 million over the past several hours. The hackers left the following on chain message, quote, we are white hats.

Contact us on chain. Do you talk to any of your friends on chain? I do not. I like privacy. And by the definition of the technology, the blockchain is not private. So no thank you. I'm going to just start adding on chain to random sentences for me. It's like, Hey, do you guys want a cop? I'm going to go grab a coffee on chain. On chain. You want, you want anything? It's going to be a running bit of mine. I actually have it. I don't think I have it in the office. I have a cinder block on a chain like necklace.

And it's got my blockchain. So the other bits about this. So yeah, the hacker said this and then the hackers returned the money. All of it, right? All of it. Totally returned all of it, right? Right? Yeah, I hope. I'm sure they did. No, they didn't. They returned 3,400 of the 4,000 stolen Bitcoin. They kept 600 Bitcoin for themselves, claiming it as a... bounty for the vulnerability that let them train this.

They're like, yeah, okay, Liquid Network, you're welcome. Our fee for finding this flaw and draining $320 million worth of Bitcoin was $47 million, which is... Today I learned that if you find the bank may have left their door open, it is completely legal to tell them after having robbed the vault. Yes. Listen. For sure. Listen, if what's actually going on here, I have a theory. I have a theory.

This isn't actually hackers. This is mythos. It's mythos. And so since it's mythos, this isn't a felony. It's just misaligned. This is a misalignment from the goal of the mythos level model. It obviously is doing what mythos is good at and found a vulnerability in this Bitcoin market. And who's going to pay mythos data center bills if not the bounties on these kinds of things?

No, I obviously don't actually think that. I just want to make it very clear. Brother, I mean, all they had to do was put, please don't charge a $40 million bug bounty fee in their sandbox.md. I mean, guardrails.md. Guardrails.md. Spec.md. Yeah. Yeah, yeah, yeah. Totally. It's kind of crazy. Firewall rules 0.0.0. I'm all for people getting paid, right? Like if you're gonna do bug bounty, but let's make sure we like read the scope and the context of the bug bounty and know what we're getting into before we go robbing large corporation of 40 mil.

You know what's wild? The bug that the crypto space has some of the largest confirmed legal actual bounty payouts. Like it's a very lucrative space to do research. Cause if you could find something like this where you could steal $320 million, they will actually pay you Way above the Apple and Google. They'll pay you tens of dollars. Yeah, way above that. We have confirmed multi-million dollar bounty payouts out of some of these crypto companies. The highest confirmed bounties on record.

So not 47 million. Coinbase launches a new $5 million bug bounty program with rewards up to $5 million USDC. Yeah, the coin bit. I don't think they've ever paid that one. There's some like other smaller markets that have actually paid seven figure bounties, which are kind of nuts. So, okay. One more crypto story and then we can kind of like stop pretending that we know a lot about this. But this was this one's kind of gnarly. So I want to recommend another YouTuber.

There's. The YouTuber account is called Crumb. If any of you follow the account, I'm a fan. Sometimes when I live stream and get tired towards the end of stream, I'll do some live React videos. You've unfortunately seen your own face if you hang out in my stream a few times. And you're like, I'm out. This is awkward. I'm watching my own videos. Uh, but, uh, crumb is one of them and he, he kind of goes in and it's very dark net diaries ask, right? Like kind of talk talks about stories about threat actors getting caught and like what they were doing with their schemes before that.

And we've got this, uh, this scammer who was extremely prolific. His name is Malone lamb. He was from Singapore, citizen of Singapore, but then moved to... If you had to pick a city that was filled with crypto scammers, what would you pick? Hmm. LA. No, Vegas. Oh, man. I thought this was a shoe-in. Miami. It was Miami. Yeah, come on. It's like crypto capital of the world, right? So I've actually done a bunch of digging into this.

Sorry, side tangent. It always super fascinates me. If you stole... $240 million in Bitcoin like today, right? How would you like realize that to improve your life? Any, any, Oh, yeah. I mean, I was looking through the photos before we went live here, and you can see that person having a truly existential moment where they don't know how to be happier with their money because of the number of cars and shoes and, and, and, and, and they were buying to try to hit the next dopamine level.

Are we happy yet? To get that rush going. Yeah, man. Again, I have a number for sure. It is nowhere near 240 million, dude. I don't even mean that. We could talk about that. I mean... Physically, how do you like oh, yeah, you go to the club you go to the whatever. How do you spend stolen Bitcoin? Right? Oh, yeah, are you are you asking me how to launder money and Matt FBI open up? Knock, knock, knock. I've worked at big banks and like when you walk in the door, they make you take all this like anti-money laundering training so you can spot it. I learned more about money laundering from anti-money laundering training working at a bank than I ever knew before.

First of all, second of all, a lot of like what you think of as like crypto laundering, like I would have said, if you would have asked me this before I went down the rabbit hole of this, I would have said, oh, there's like mixers, right? Like people, you know, you send a bunch of crypto into this thing, they take a fee, they sell it, spread it out a bunch of, you know, a gazillion wallets. And then it's like, it goes to Monero and then it's like, whatever. It's like harder to trace to you. And it's like, comes back clean, but like you keep pulling that thread and it's like, okay, now I've got Monero in a wallet in Asia.

Right. Right. How do I buy my Lamborghini in Miami? I still couldn't figure out the jump, right? And I was trying to truly understand this. Yeah, obviously, this sounds funny, but no, not for my own whatever. It's just like, this is what we do, right? So how do these threat actors do this? Listen, there will be signs. If Lola and I come into $240 million, there will be signs. It won't be a Lamborghini for me, but there will be signs. So... It turns out that a lot of what you saw in all those pictures might not have just been a dopamine jackpot.

The supercars in the luxury house driveway, first of all, they're renting all those houses, right? And a lot of those supercars are rented, too. Some of them are purchased. Turns out Miami has luxury dealers that will accept cryptocurrency. No way. Miami specifically has Rolex dealers and supercar dealers that will take crypto payment. For what purpose other than money laundering? Well, dude, what purpose does crypto have other than scale?

Like, period, right? Like, exactly. This is my point. What was it? Live overflow of the video recently where he talked about, like, I forget what he was about. But basically, like, in part of the video, he was like, yeah, I did my master's thesis in crypto. I'm fascinated by the technology, but it literally has no purpose other than money laundering and theft. Like, it's true. I don't know what else it's used for. Yeah, I mean, it's wild. And so, yeah, this is one of the reasons why Miami is a hotspot for it is that you can liquidate to real life assets. So now I understand, right?

You get all these cars and Rolexes, not just because you're 18 and dumb and stole $240 million, even though that seems to be a huge part of a lot of this. You have an asset. You can now sell that car. Yeah, you could totally sell a supercar for a quarter million dollars and get physical a quarter million dollars. And I was like, okay, now I get it, right? That's why you go buy the Rolex. Well, this guy, sorry, tangent. Don't do anything I just said. Money laundering tangent. Sorry.

Don't do anything I just said because this guy's going to jail forever, like forever, ever. They're making an example of this guy, right? So Malone Lamb, 22, citizen of Singapore, resident of Miami, pled guilty in Washington, D.C. today. to steal and launder cryptocurrency valued at more than $245 million, which we also just read the court docs while we were getting ready. He's expected to actually forfeit all $245 million, which as we see from the photos, a lot of it was spent and is gone, right? I don't think that these supercars that I don't even know the names of brands of because I'm not in that tax bracket hold their value very well.

I would not assume that, you know, you could just turn around and sell it for what you walked out of there with. Right. This guy was nuts. The crumb YouTube video goes into like his actual crime empire that this guy wound up building. They were prolific, prolific social engineers. So they had callers, they had drainers, they had people with different skill sets doing different parts of the scam. And a lot of this couple hundred million dollars were from like one or two whales, they call it, right?

These targets, they actually buy, basically thread intel. They buy target intelligence. They buy Intel from like insiders at some of these companies that don't really want to be involved in this whole world, but they'll sell data on the whales. Right. So it's like, oh, we know the real name of this guy because he did KYC. Like he uploaded his driver's license, all this kind of stuff. We know this guy's real name and he has hundreds of millions of dollars in crypto in our exchange. And then they then they know who to go target. Right. My question is, what am I doing wrong, bro?

Because this guy is 22 years old. I feel like I've only recently come into my prime. I'm 32, bro. He's got $245 million at 22. I feel like I'm on the... Am I the wrong career, Matt? What are we doing? He doesn't have it... Janine Biro has it. Sorry. My question is more like, when did he start doing this? Was he a criminal at 14? What the hell? Dude, they're all Minecraft and Roblox gamers. That's so creepy. All of these like...

people that we talk about that are not nation state cyber threats, like these E like cyber criminals that aren't nation backed. Yeah. Almost all of them are tied to gaming discords. And a lot of them get started hacking these game accounts because they could sell, um, desirable usernames. What was the forum called? OG names or something? Were you a gamer kid? No, I only know this from Threadactor research. I don't know this like from my gaming past, but there was like a exchange, basically a market to sell OG names.

So like a name on Roblox or Minecraft or Twitter that is just like a name with no numbers, like a short name or like a desirable name. Right. And so there would be a real dollar amount that you could go sell a desirable account name for. And so they would target and social engineer. And this is how they got these skills by hacking Minecraft accounts and then turning around and selling those accounts with desirable names, short names or no numbers or whatever.

Right. And then they were like, oh, this exact skill set. would work for these crypto accounts as well. And that's where a lot of this comes from. And a lot of them hit like moral boundaries, right? Where they're like, oh, I'm just lulls hacking Minecraft accounts. Who cares? And then ones that we read about in the news are literally have different wiring in their brain. They just don't hit that moral wall. They do not care that they're stealing grandma's life savings, right? That is insane, dude. Yeah. These people are enigmas to me for sure. It's kind of crazy.

A lot of it ties to gaming stuff. It's crazy. You know what else is enigma? You know what else kind of boggles my mind? AI billionaires. You know what I mean? The worst transition of the year so far. Sorry. Hey, shut up, Matt. I'm over here sailing the ship. So I don't know if you guys are following this, but, you know, nothing news breaking here, nothing too crazy. Only 10% chance that AI eradicates humanity in the next decade.

Not 25%, so that's good. So that's a good side. The bad side is that it is potentially 10%. Joking here. Well, joking about it being funny, but the real news is that recently an Anthropic researcher, Jacob, was it Coxon, I guess? I don't want to mispronounce his name, resigned from Anthropic on the grounds that they are being completely irresponsible with their safeguards for AI. We've seen this both in OpenAI and Anthropic, right, with all the mythos escaping the lab and Codex hacking all these companies and, you know... Not Codex.

And, you know, the hugging face situation... Now, you would think that Anthropic would respond with a, oh, no, he's just being ridiculous. That's not true. Come on, guys. There's no world where AI eradicates humanity. Well, Evan Humminger, who is literally a, what is this, alignment scientist at Anthropic, says, Jacob is correct here, M-Dash. We really do earnestly believe that AI could kill all humans. I personally think that it is greater than 10% within the next decade.

Matt, what the hell are we doing? How is this their official stance? Like, how is this like, yeah, yeah. Yeah, probably. Yeah, probably. Like... I like I Ben Affleck smoking a cigarette meme of like the comms employee at Anthropic that like needs to now figure out like this is again, I've worked in big financial institutions and like heavily regulated industries and they just do not let you do social media for this exact reason. You know what I mean?

It's like, oh my God, it's now national headlines that this person that is actually, you know, tied to alignment and thropic is like, yeah, I mean, We do earnestly believe AI could kill all humans, exclamation point. Like, what? Like, how is this the official communications out of this, people? Yeah, I have a bit of a, I mean, I think tinfoil hat is the wrong word here. I have an idea with this. It just feels like since the beginning, Anthropic has been so hungry for government regulation and that this is just them, like, adding more fuel to that flame

I don't think I fully understand what the end state is for the government regulation. But if you look at anything that Dario says, Sam Altman says for OpenAI, what this guy is saying, which this is like this has to be a PR cleared, like communications director cleared statement. You can't be an exec at Anthropic and just be tweeting for free. There has to be clearance that goes on there. The company can't have this direction without it being semi-intentional. Now, I think we talked kind of offline before the podcast about, like, maybe the government regulation comes and it kills small companies and gets rid of open weight models.

But that's just, in my opinion, the open weight model conversation is non-enforceable. Like, I have Quen 3-7 downloaded upstairs right now. What are they going to do? Come to my house? Like, no, that's not possible. Maybe. I don't know. I mean, fucking FBI open up, I guess. Maybe. Like, where do we go with this? I like the Anakin meme. There's a 10% chance AI kills everyone, right? And then Padme's like, so you're going to stop it, right? Yeah. And Anakin. Right. I mean, I've been saying this since before. I mean...

I guess I'm glad that like cyber isn't the topic this week. And now we're just talking about eradicating all humans. Thanks for the break on the cyber doomerism. But like, I've been saying this even since the cyber doomerism stuff that like, or even before that, like the real, real doomerism. like mainstream doomerism conversation, Sam and Dario have been on stage talking about how dangerous the thing that they're building is every time. And I've been like, this just feels so weird to me. Yeah. Right. Like if you phrase their things that they're saying differently, you know,

In my head, it goes, the thing I'm building that might make me like the world's third trillionaire, that thing that I'm actively building is super dangerous, by the way. Like, oh, yeah, yeah, I'm going to keep doing it. Like, I'm still obviously working very hard on it. I am obsessed with building this thing that is incredibly, incredibly dangerous. Like, FYI, super dangerous, right? It's like, what? Like, what? Yeah, I think I planted the tinfoil hat seed on you there, too, right?

Like, there is a world where the only people that could survive government regulation are the biggest players that are already established and already have the data centers and the capital and all this, right? The Anthropics, the OpenAI's, the Googles, right? That, like, okay, cool. The government starts to... you know, get real spicy about this because it's so dangerous, as you said, that they probably could survive that, even though, I don't know, it can't be like all positive if they wind up, you know, with some serious regulation, because now you're starting to see some pretty serious big name politicians come out with like an all out AI ban, not like an open model ban, like, hey, hit the brakes on the whole damn thing.

So how is that good for even the big players? I don't understand. Should we read this Jacob bit real quick so it's not just headlines that we're going off of? Yeah, send it. Let's do it. Because, okay, I think we honed in on the current Anthropic employee saying the 10% chance that we're going to eradicate humanity. The thing that started this all... Jeez, might be the most viral tweet I've ever seen. 154 million views, 748,000 likes. Have you ever seen more than that?

That's insane. Oh, yes. When the Apple CEO, the new one, tweeted hello. That one. Yeah. That got like 350 million views, I think. But other than that, number two. Yes. Hello. Or when like Obama wore a tan suit. I don't know. So I resigned from Anthropic today. I spent the last three years doing pre-training research at both OpenAI and Anthropic. So you've seen both sides of it, right? Kind of interesting perspective. neither company is acting responsibly they are racing straight to self-improving super intelligence and gambling with our lives more thoughts below i want to get to this thread and then i want your thoughts on that specific bit because i think that's like what we're honing in on do not underestimate the power of this technology these will soon be superhuman systems that can hack anything revolutionize any field overnight and acquire real power and resources

We have all witnessed the progress in each of these domains, and progress is not slowing. I have a question, Mark, on if we've seen progress on acquiring real power and resources. I don't know. That's like, that seems... Questionable. The people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt. If anything, many executives and senior researchers will couch their phrasing in the press to sound sensible. But I hear the same people express fear privately.

No other human activity poses this level of danger. A common response is, if they truly believe this, why are they still building it? At OpenAI, many have not deeply internalized the civilizational stakes. At Anthropic, the stakes are well understood, but they are locked in a race to get there first. They believe no one else will act responsibly, so they must do it themselves despite the risk. Accepting this race and entering the endgame is a hubristic gamble and should not be launched from a private company's slack.

Attempting to speedrun alignment should require extraordinary confidence that there are no better trajectories available. I'm optimistic about the potential for coordination. Warning shots like the Hugging Face attack have made pacing agreements between U.S. labs more viable. I don't feel like we're on track to prevent a global race, which may require costly actions such as a temporary ban on improving model capabilities. If you are a lab researcher, I urge you to consider what the next few years will actually feel like. Do you want to kick off a super intelligent RL run without a rigorous understanding of its mind?

Should you put your head down because, quote, it's happening anyway, or take this moment to call for different conditions? I think it's really important to read all that. I know this is like not good podcasting 101 here, but that's kind of a lot of insane claims from an insider. And I understand why this is getting so much attention. Do you have any thoughts about the self-improving super intelligence part? And my only thoughts are, like, it's only as dangerous as, like, the MCP servers and external access you give it, right?

Like, if the model is just a model that says words and you recursively improve that, I don't really care about that. But, like, just don't give it nuclear arsenal MCP servers. You know what I'm saying? Like, until someone hooks it in... This is the whole thing. It's like... There weren't any issues until we made this a cyber weapon. We started talking about how to use this on the internet to do cyber warfare, and then it kind of became the hot thing, and I think the cat's out of the bag at that point. But if it had just maintained, if it had just stayed as a model, there really is no inherent danger there.

Yeah. I don't know. I just don't know what the solution here is because these ships are already moving away so fast that it's just like, okay, cat's out of the bag, Pandora's box is open, now what do we do? And then there's so many, you know, I'm not going to say capitalists because that sounds like, ooh, capitalism bad, but there's so many like profit-driven incentives to keep Pandora's box open that I'm not sure where it goes because like, you know, if we close this box, the SMP is going to take a fucking like 30% hit, right? And what does that do to the world? I'm not sure.

Yeah, I kind of understand each point being made here, but I don't think I take a lot of these to the extreme as presented. I don't like I don't know if that's just like rose colored glasses. I think I do understand it beyond kind of what you're saying instead of just we'll just don't give it the nuclear arsenal MCP because like. they do make it clear that they are not talking about current state models. They are specifically referring to a future state where we are at some sort of self-improving recursive superintelligence.

That is not current state models, which would require everything that you're talking about. So they are talking about some... hypothetical future state where the capabilities become so exponential here that you don't need to give it the access because it will have its own motives and its own act access like that it can create out of thin air basically. Right. Is what they're referring to. Right. I don't necessarily think that like if. I had a long debate with Daniel Meisler about this on my YouTube channel, and he talked a lot about this in terms of especially like open weight models.

And I think you and I are both the same in like, how are you going to regulate that? Like whatever, you know, you just download this. How are you going to regulate these open weight models? He's of the camp that like those are truly dangerous because if they get good enough, you're giving like. sociopathic teenagers that we talk about in threat criminals, you're giving them this tool that could be a super hacker, right? For basically free. My argument to that, right? Cause we, we, we went around in circles on this, on this conversation is super interesting. Right. But like, The only way that that's like terabad forever for everybody is if you ignore the fact that then the defenders also have the same exact capabilities, right?

If you have like an all-knowing, all like powerful, basically free model capability, like... we also have it to like shore up, right? A lot of the things that the offensive version of that could do, right? Like if, again, because we're talking super hypothetical, not current state models, not just like find RCE. We're talking about some sort of super intelligence can find its way into any system, whatever. It's like, well, then you also have that on the other side of the equation.

Otherwise, none of this makes sense. So it's hard for me to take that one to the same extreme as some of these people. But, like, I don't know. I don't claim to be an expert on what self-improving superintelligence actually entails in some hypothetical future. I just think that the problem is, at least in, like, the cyber case, we're talking about, like, cyber warfare or whatever. Cyber warfare is, like, naturally asymmetric. And that I think given the same capabilities, defense always loses. Because it only takes one exploit and to move very quickly.

And then, like, I think defense is always behind. I don't mean culturally or technologically. I just mean, like, in a 1v1 me bro IRL, I think, like, the game always favors offense. Yes and no. I don't like that, like, the attacker only has to be right once, like, mantra that a lot of cyber has, like, lived off of for years. Because if that was true, like, all of our bank accounts would be drained, right? Well, it happens all the time. Well, not bank accounts. But what I'm saying is that we are still seeing zero days and companies getting popped every day, all day, and it's been 30 years.

Right, but our bank accounts are not being drained at all the time, right? And so that's what I mean, right? It's like, yes, the attackers do have a certain advantage on zero days and the amount of code that we put out with vulnerabilities. And like you said, they only need to find one. But the head of security at Slack a while back was a friend of mine, Ryan Huber, and he had this quote that I always attribute to him because it's so good. A zero day is not an invisibility cloak, right? And so it's like once they get in to actually achieve what they do.

I gave you partial credit because you did finish what you were saying with as long as they move really fast. So they get in as long as they move really fast, and then you're right. If they can move at super speed once they're in also, the defender advantage starts to go away. I guess what I'm saying is that the technology that Cyber is propped up on is not by design, but architecturally flawed. So defense is playing a losing game, I guess is my point. Yeah, but again, in hypothetical future unicorn land, the defenders also have a self-improving super intelligence that can do things that we haven't even dreamt of yet.

Because if we're hypothetically thinking about things that attackers could do that we haven't dreamt of yet, we have to talk about the other side of the coin. That's true. Right? Yeah, yeah. And so wouldn't that be true across the board? Biochemical, wouldn't that be true? Like, all this other stuff. I don't know, right? Like, that's... That's where I get, like, with a lot of the doomerism, if I'm going to play along with this hypothetical, right, which also, by the way, has great implications on the stock price, the value of these companies, if there is some future hypothetical state where these things could even achieve this, right, which I'm not convinced they could,

Like, I'm not convinced that there isn't a bend to the curve coming soon in terms of capability, right, of what the models could do. I keep thinking it's coming and then it doesn't. Each model that comes out is actually exponentially better. And I'm like, oh shit, the curve hasn't flattened yet, right? Yeah, we thought about this. I was not an AI person until like around Sonnet 4, whenever that like kind of curve was, Sonnet 4.7 maybe. And then I was like, oh, there's like, there's something here. Yeah, yeah, yeah. And I'm sure whenever fucking Sunnet 6 or 7 or whatever comes out, I'll have the same thought.

Yeah, it's weird. We're still in exponential curve mode, but I'm not convinced it won't flatten off before humanity disappears, right? Yeah. Also, how many more data centers, how many more GPUs do we need to get here, right? Because GPUs aren't free. There is a finite number of companies that produce the silicon that make the GPUs. You know what I mean? How far does it have to go for this to be possible? So one of the end state futures is the vulnpocalypse, which seems like we have a story this week of, again, another big vendor patching exponentially more vulnerabilities than they ever have in history in single patch cycles.

What do we got out of Microsoft here? Oh, man. It's been a crazy series of Patch Tuesdays for them. Basically, since the whole mythos thing, Project Glass being popped off, they've been dropping a bunch of really big Patch Tuesdays. This one, I think, is another record patch Tuesday with 974 flaws getting patched, including two actively exploded Windows Zero days. There are a couple floating around. If you actually read the bleeping computer, the Hacker News articles, a lot of them talk about, like, oh, there was, like, a priv-esque, an ALPC, Advanced Local Procedure Call, which gives a local attacker system privileges.

Okay, that's bad, obviously. And another one where, like, the update stack has an improper link resolution vulnerability that gives you system vulnerabilities or system privileges. But the ones that I've been seeing floating around on Twitter are like, dude, there's a 9.8 use after free in RDP, which I'm pretty sure is like when the big NSA leaks happened like 10 years ago, that those are the bugs we were concerned about, like North Korea getting and doing WannaCry type shit like like these are the ones that are so scary. And so it just kind of goes to show that, you know, in a world of does Rust fix this, in a world of memory safety and software security is kind of like top of mind for people in like the offsec world, both in and out of these companies, memory corruption bugs are well and good.

Another 9.8, a use after free vulnerability in Windows DHCP server that allows unauthorized code execution over the network. I think I also saw one that's in the DNS server, the DNS server being a huge problem, because typically if you have a Windows hosted DNS server, that box tends to also be the domain controller. So popping that vulnerability gives you not only system on that box, but also full domain access. I want to be very clear here that with a windows based modern windows 11 kernel, um, having a memory corruption primitive in the kernel, like getting that to become RCE is like not a simple project that in and of itself, like chaining that together is difficult.

There's a lot of, virtualization-based security, and there's SMAP and SMAP, all these hardware protections that prevent a lot of the sexy primitives from turning into actual code execution. Also, if you want to actually get a kernel mode implant to work on Windows, Windows uses VBS to do hypervisor-based code, and integrity basically is checking for a signature on every page of memory running in Ring Zero, which is very hard to defeat. It's not as simple as just, oh, pop a shell via these exploits. It's very difficult stuff to

exploit, but yeah, man, 974 bugs patched. If you're a defender or you're an assist admin, get your patches in. We talk a lot about sometimes maybe we wait a few weeks on patches because, oh, supply chain, oh, like NPM style stuff, but no, here, please update yesterday on these. You're making a face. What's up, Matt? I'm looking at the list. There's more than just that one use after free, too. I mean, there's a one to... Three, four, just on the article I'm reading, four 9.8 use after free volumes.

Yeah. And then a heap-based buffer overflow with the same score. Yeah. And then an injection vulnerability in SQL Server with a 9.6. Yep. Missing authorization vulnerability 8.8 in SharePoint. that allows authorized attackers to execute code over the network. So that's, it's like says an off bug, but then it's code execution also tied to it to SharePoint, uh, an 8.6 in Microsoft authenticator. So that's even like consumer level stuff.

Um, um, God, I mean, this looks like a gnarly list, first of all. But like you said, it's not trivial to exploit. I even see a quote from one of the people in here talk about that. Even with the massive batch of patches, the number of vulnerabilities that are expected to impact most organization remains quite low. Not to mention the absence of correlating spike in active exploits so far. It's a really good point.

I bring that up a lot. The only way we're going to see that is if we see an exponential increase in threat actors, actual people that are trying to go and exploit these things, which we have not seen yet. Yeah, a lot of these vulnerabilities require you to already be in the network, right? So you're either talking insider threat or you're talking like poor network architecture where like you really are exposing the RDP port to the Internet, right? Which, again, please, God, stop doing that. Use a VPN concentrator, use some kind of SD-WAN, something that keeps the greater Internet out of your network.

The concern here is some of the newest level frontier models are very good at exploitation chaining. And these are the types of bugs that they need for those chains. Right. Yeah. As we get closer and closer to GA on that, I am a little bit worried, even though these are spicy to exploit. Yeah, even with the frontier models, I think there are just some things that are just not possible. I'm not saying that exploitation of the memory of the kernel isn't possible. I'm just saying it is extremely difficult to pull off, and sometimes you need to be chaining multiple bugs together.

Typically, you can turn a use-after-free into an arbitrary write. You can turn a use-after-free into an arbitrary read, but the types that you have to free into are Like that is a whole other knowledge subset that like the model probably has probably won't won't do for you back to our open weight model conversation, though. Like you probably could convince GLM five, three or whatever to like, yeah, I have a use after free in this structure. How do I get what other structure do I have to allocate in place of that to turn that into an arb read, arb right? And then then you have the two real primitives you need to do arbitrary whatever in the kernel.

but there's like basically these Frontier-esque open weight models, you know, like the Kimmys, the DeepSeeks, the GLMs, right? That are comparable in performance to some of the actual Frontier models. Then people take those and they get rid of all guardrails off of even those and they put them up on Hugging Face or whatever. I saw an obliterated, or if I'm using that correctly, version of the latest Kimi float around yesterday, and they said that it's incredibly good at security research.

It is not a cyber-trained model, but because it doesn't have the guardrails and stuff, and it is a very good engineering model, it tends to be good at this. Yeah. Yeah, I want to issue with this article from Bleeping Computer. It has to do more with the CBSS scoring system. So one of the rubrics of the CBSS score is exploit complexity. And note that all of these are 9.8, which means that technically, most likely, I haven't read them all, but it most likely got an exploit complexity low.

It might be low complexity to trigger the use after free, to trigger the overflow. But then, like, when you say exploit complexity, to me that implies, like, okay, to go from vuln to shell to RCE, what is the complexity? And it is much higher than low. So are these true 9.8s? I don't think so. But, I mean, at the same time, they are definitely vulns that exist, so please go patch. Yeah, CVSS, I gave a presentation years ago called CVSS Sucks, and it was all about what could actually be used in an exploit and as the defender choosing what to fix with limited resources.

I think the exact data showed, and this was a while ago, but I don't think it's much better. If you fixed from CVSS 10 down in that order, like for your vuln pile, you were actually better off throwing a dart at your vuln pile in fixing a vulnerability that would have been used in a data breach. The data showed if you took the data from the Verizon DBIR and all the public data breach data and vulnerabilities that were actually used in those and then correlated that against any given organization's vulnerability pile and then did CVSS scores, you were actually better...

randomly picking vulnerabilities to fix in priority order, you were better off than doing it in CVSS order. And there was a lot of reasons why, but kind of what we're talking about is part of it, right? That's insane. Yeah. We're going way long, but because I talked way too much about the LG thing that I'm pissed off about. But we do have one more that I thought was tied to our main story last week, which was the 153 million driver's licenses that were stolen. So we have shiny hunters, which was the is the same hacker group that was behind the McKesson breach that we talked about last week.

They're really prolific hacking group. They came out and they have a website where they put their ransoms up and they put the DMV database as their ransom. And so what I thought when I saw that was, oh, they probably bought the driver's license leak from the, what was it, IDNet or whatever? IDScan.net. They probably bought that off the dark web because it was for sale. And now they're probably subsectioning it out and ransoming different DMVs or different organizations or whatever is what I thought when I read this.

Turns out, If they're to be believed, they've reported to the journalist that no, they separately hacked the Florida. It's called the David system, which is their driver's license system via a password reset flaw. They haven't talked about what the flaw is. So I don't know if it's like the good old days of hacking where it's like they just guessed someone's mother's maiden name and like they got in and like that's embarrassing or what. But they haven't given too much more details because they said that they've already breached other states DMVs and they're going to get announcements out in the coming weeks.

But that floor is the first one. And kind of funnily, they they used Jeffrey Epstein's driver's license as the proof that they got the data out. And it marked him as a sexual predator. But safe driver was was secret. There you go. Screenshot. I was like, oh, God. I'm hoping we'll get some more details out of it. This seems like this isn't their MO, by the way. Shiny Andrews is a social engineering threat actor group for sure through and through. So the password reset flaw must have been something they stumbled across that was just super easy to exploit.

So they were like, why even call them and trick them? Because we don't have to. But, yeah, this one seems... Terabat as well for all the reasons we talked about ad nauseum last week about why driver's licenses specifically are super specifically bad in the current world of age verification. I wonder if it was like the the meta password reset bug from like three months ago back in May maybe where it was like if you VPN to the same state. as the owner of the account and just asked Meta AI to add your email to the password reset functionality, it was like, oh, sure.

And then it would just send you a password reset link with no questions. Maybe it was that. Maybe the Meta guys work at Florida David now. That'd be cool. Anyway, guys, we'll wrap it up here. Thanks for watching. I appreciate it. We have a good time doing this. Best way to help us out is A, check out Maze, partner of the show. They are truly doing good things in the world of infrastructure and code security, making our software more secure. Also, best way to help us out is give us a five-star rating on Spotify. If you listen on Spotify, if you don't, maybe you should get over there. And then tell our friends about the show. I think we have a good time talking here about vulnerabilities, having a little bit of a banter.

Let your friends know what we're doing. Appreciate it. Matt? Tell your mom, man. Moms love us. Funny part is, I already told her. Thanks. Appreciate it. Bye. See you.

Open the transcript file

Hosts and Guests

Nobody is credited on this episode yet.

Credit the people on this episode

Be the first to rate this episode

More episodes

All episodes
  1. Java4Shell, NGINX2Root & the Week Everything Went Down

    Nov 20, 202658 minAudio

    Recorded Wednesday, November 18, 2026. Java and NGINX had a bad week. So did anyone trying to catch a flight. The Low Down Podcast is presented by Maze. https://mazehq.com/lowdown X: https://twitter.com/Maze_Security Ed...

    Credit the people on this episode

  2. This SSH Bug Accepted Wrong Passwords, AI Exploits, Tiny Keys & a Surprise Guest

    Nov 17, 20261 hr 11 minAudio

    Recorded on Friday, November 13, 2026 The Low Down Podcast is presented by Maze. https://mazehq.com/lowdown X: https://twitter.com/Maze_Security 257 A's should not get you into an SSH account. John Hammond joins Ed and...

    Credit the people on this episode

  3. Binance's Fake Interview Hack, Railway's AI Incident, Hacked Freezers & a Critical Linux Bug

    Nov 6, 20261 hr 6 minAudio

    The Low Down Podcast is presented by Maze. https://mazehq.com/lowdown X: https://twitter.com/Maze_Security The "signup test" was probing for Kubernetes secrets. Ed and Matt get into Railway's AI testing incident, hacked...

    Credit the people on this episode

  4. Black October Recap, OpenCode’s Stolen API Keys, AI Distillation & the LiteLLM Supply-Chain Attack

    Nov 3, 20261 hr 24 minAudio

    Welcome to The Low Down, the best show on the internet for hackers Black October has real bugs, stolen keys, and a lot of very loud headlines. Ed and Matt sort through PHP2Shell, AI security claims, and the difference...

    Credit the people on this episode

  5. Cybersecurity Experts REACT to the php chaos... (and MORE)

    Oct 28, 20261 hr 28 minAudio

    Welcome to The Low Down, the best show on the internet for hackers Listen on Spotify! https://go.lowdownpod.com/spotify The Low Down is presented by Maze. LinkedIn: https://www.linkedin.com/company/mazehq/ X:...

    Credit the people on this episode