← Cybersecurity Headlines
Cybersecurity Headlines
Department of Know: CISA's cryptography categories, Gottumukkala's ChatGPT gotcha, NTLM says goodbye
00:46:00
Loading summary
Transcript105 lines
- [00:00]
A
This is Rich Stroffolino with the Department of Know. Steve Zaluski, co host of Defense in Depth. I have to ask, what is your priority this week?
- [00:12]
B
The audio and video social engineering problem. And we're going to talk about that.
- [00:18]
A
Oh, yes, yes. Top of mind, I think for a lot of people. Yeah, it poses, it kind of challenges everything that we're seeing and test so many assumptions.
- [00:32]
B
Right.
- [00:32]
A
That's, that's what excites me about that. So I love that you're thinking about that, Steve, But Nick Espinosa, host of the Deep Dive radio show, I got to ask you the same question. Where is your mind at this week? What is your priority? What are you focusing on this week?
- [00:45]
C
Well, right after this, I go and actually to my own radio show where we're going to be talking about protecting yourself in a surveillance state. I wonder why. You know, wonder why. Wonder why should be a good one.
- [00:57]
B
We're going to talk about a little.
- [00:58]
C
Bit of that today, too, so I'm excited for that.
- [01:01]
A
Well, yeah, I'm glad we can kind of give an appetizer for something you're gonna be talking about a little later on. I love that little bit of synergy. All right, producer Steve, let's run that opening and get into the show. From the CISO series, it's Department of no. Welcome to the Department of no. Your Virtual Monday strategy meeting. Our sponsor for today is strike 48. Agentix Security. Start moving at machine speed. Remember to get involved in our YouTube live chat. I see some familiar faces in there, including Kevin Farrell. Find the true two. And the big boss man, of course, David Spark in there having some fun times, making sure everything is working as it needs to be. It will be, rest assured. So get in there and join us. We're live Every Monday at 4pm Eastern. Or email us feedbacksoseries.com Just a quick reminder that the opinions expressed on the show are in fact those of our guests, and I'll even say me, but not necessarily those of their employers. We've got about 30 minutes, so let's jump into it. Starting out with our no or no segment. This is where we have so much news to cover. We, we need to know, is this a story we need to be paying attention to bringing to our teams talking about at work? Or is this, hey, no, thank you. Not a big deal. First up here, CISA releases new cryptography categories. The story is about CISA's initial list of hardware and software product categories that already support or are transitioning to post quantum cryptography. That were developed with the NSA under a 2025 executive order. CISA says future purchases for cloud, endpoint security, messaging, and so on should be PQC capable to prepare for Quantum era encryption risks. So, again, I'm going to say this is another brick in the wall of, hey, Quantum's coming. We all know it here. So I got to ask, is PQC a no? You know, do we need to be talking about this more? Do we need to know about why this is on the horizon, or is this a no? Thank you. For you, Nick.
- [02:51]
C
Yeah. So this is definitely a no. And you can figure that one out.
- [02:56]
A
Come on, Nick, help us out.
- [02:58]
C
No as in knowledge. I think this is actually a good one. And to me, the way I look at it is CISA is clearly indicating a shift from just being like a research topic to actual procurement reality.
- [03:10]
A
Right?
- [03:10]
C
This isn't theoretical guidance that they are putting out. I think that this is like signaling, like an early market standard that post quantum readiness has to be the baseline. And that's what it is. I mean, if you think about it, we are potentially delaying future breaches. So if somebody goes in and steals your encrypted data. Well, if it's not, you know, if it's not quantum resistant, then maybe in two years or whenever they can crack it. And now we've got a breach that could have happened two years ago or today, and it's happening in the future. So I think this is a. This is one to definitely talk to the team about.
- [03:44]
A
Steve, what about you? When. When Quantum gets a speed type, right? Is it something we need to know more about, or is this something where you're saying, interesting, but I'm not bringing this to the team yet.
- [03:56]
B
I. To me, this is a yawn. Okay. It was a slow. It was a slow news week. And Quantum is one of those things that every couple of weeks or month, right. It brings itself back up because people want to sell the problem. And I'm like, great, another chocolate chip cookie recipe. Awesome. Okay.
- [04:25]
A
This is. This is the recipe website that. Oh, my. I found a new recipe from my grandma. So we have to. That is. That is interesting. All right, But. Okay, so, Steve, if this is not. What would be. Are we just waiting for the Z wave or whoever? Google comes out and says, we have a production system. We're breaking encryption. Like, is it. What is the. What is the. The signal for you that you're looking for that would make it a non. Yawn.
- [04:50]
B
So this isn't the problem, right? The problem exists in that we have a bunch of old cryptographic algorithms that are out there that are. We call legacy that we can't upgrade if we can push those out. It's relatively simple to deploy post quantum choices for the asymmetric encryption. It's just a matter of we got a whole bunch of applications that can't be upgraded. So no matter what you tell me here, okay, it can be the best chocolate chip cookie in the house. Okay, but if I'm allergic to chocolate.
- [05:26]
C
Are you telling me I've got to get rid of my MD5 hashing? Is that what you're telling me?
- [05:31]
A
You know how easy it is to recover my chocolate? Hey, hey. We've got comically old things later on in the show. Let's not lead too long, all right? All right. I like the agreement to disagree here. It's very fascinating. Next up here, US Cyber Chief uploaded sensitive files into public chatgpt. According to Politico, US Acting Cyber Chief Madhu Gutamukala uploaded contracting documents marked for official use only into a public version of ChatGPT last summer. The documents were classified and he had received special exemption to use ChatGPT at the time when it was blocked for other DHS employees. Thanks. Office of the ciso. With companies struggling to establish and enforce correct behaviors in using LLMs and these, one rule for me and another for you, positions also being demonstrated. I don't know about the UK government. We could choose an organization. You'll probably find an example of it. I don't think set the best example. I'm curious, Steve, for you, is this know a little more or a no, thank you story for you?
- [06:31]
B
So this is me being 21 years old and coming home and my father drinking a beer and telling me you shouldn't drink because it's bad for you. Do as I say, not as I do, okay? I'm like, granted in this case, okay, It's a little blown out of proportion here because it comes across as do as I say, not as I do. Okay? I get to do it, and I'm going to tell you how to do better. And it kind of rubs everybody the wrong way. But that aside, look, we all make mistakes, okay? Whether it was a mistake or not, I'm like, we're going to continue to do this because to a certain extent, we reserve the right in management to do what we think is right, not just follow the rules.
- [07:17]
A
Nick, what about for you? Where does your mind sit with this big deal? Not a big deal on this one.
- [07:22]
C
I mean, come on, dude, really? Like a Senior official within the Department of Homeland Security is using a public AI tool and just bypassing standard data handling controls, despite knowing its sensitivity for official use only has been around forever, aka Cui.
- [07:40]
B
Why?
- [07:40]
C
FCI, take your pick. Like, this is Governance 101, man. And so the way I look at this is like, okay, if I'm talking to a stand up about this, just bring it up to reinforce. This is why we have policies and procedures, right? And you're telling me that the Department of Homeland Security doesn't have an AI acceptable use policy this guy signed? I mean, I kind of question his credentials.
- [08:03]
A
I'm not going to lie.
- [08:04]
C
I don't know the guy. I'm sure he's a nice guy, but like, how do you, how do you make that mistake in that position? And I deal with a lot of this kind of documentation, you know, and these markings quite a bit. I've never made this mistake. Like, I just, I just don't get it, you know.
- [08:19]
A
Well, yeah, that's, that's where it is for me because there's, there's a lot of reasons why I could give slack, right? It was last year, like, we all forget how fast all of these tools are moving and how quickly we've integrated them into everything we're doing. So it's like, all right, maybe a little slack there. Maybe this, maybe, you know, he's the acting head of CISA with, you know, maybe someone coming in from outside industry that came in isn't used to, you know, particularly red tape around the government. But no, he's actually been working in state and local government for quite, you know, his long history in that regard too. So there was a lot of things that, like, like some egg on the face. But I, again, if you could bring this to your team and be like, all right, here's how we can do better. We can be better than the acting head of csat.
- [09:01]
C
Right, Right. And that's the reinforcement point. But to your previous comment on how fast AI is moving, what hasn't changed since the dawn of the ChatGPT era in 2022 or so is that they consume everything that you give them. If it's a free and public system, you're the product. I mean, heck, Facebook taught us that, right? Google did. So I don't care how sophisticated the AI gets, if it's a publicly used system, you should know better, especially as a government employee. I've never been a government employee, but I, you know, I've worked with all of this kind of data on behalf of their contractors and government entities as well in my, in my career history. So I just don't see this. I don't, I don't see any excuse for this. You know, I mean, was it glug, glug, glug, use chatgpt or you know, like, like, I just, I don't get it. And they have their own internal systems that they've spun up. They've got internal LLMs. So like, I just, I don't, I do not understand this one. It just doesn'. It's just lazy to me, you know, I'm sorry.
- [10:01]
A
Well, maybe we could Vibe code away to a better solution here for cisa. Oh, wait, that's the story here. When Vibe coding goes wrong, there's a new ransomware strand called Sicari that is so poorly built that paying the ransom won't decrypt victims data. Some could say that's a feature, but we're going to say it's broken. The malware generates fresh RSA keys on each execution and discards the private key, leaving no viable recovery path at all. So I have to ask, when we see these types of stories here, you know, we're all terrified of, you know, AI enabled malware, right? We can write this at scale. Does it become more dangerous when it's poorly, seemingly poorly written for the purposes of a ransom? Nick, when you see these kinds of stories, where does your mind go?
- [10:47]
C
All right, so for me, this is like a brief. No, like, in other words, like briefly explain this one. I mean, just, my God, like, AI really is a tool for lazy. And don't get me wrong, don't get me wrong, you can use it, you know, if you're productive too. But I just think we're going to see an increasing volume of this and it just is going to lead to unpredictability. You know, it's. This isn't a sophistication issue, it's Vibe coding, you know what I mean? But, but it's going to basically, I think, inject a wild card into this. Because if you can't pay to recover, you know, like, and that's the assumption that we run on if we are in an incident with a ransomware event, that we have that option. Although we're not supposed to negotiate with terrorists, am I right?
- [11:30]
A
Right.
- [11:30]
C
You know, but it happens. And now you can't do that. If it's just blowing away its own private keys, then, you know, then everybody's stuck until the quantum machines come out and we can just crack them, you know, so.
- [11:44]
A
Yeah, Steve, what about for you? Like, is. Is this the. Is this how we get people to stop paying the ransom here? I'm curious, do you want to know a little bit more about this or is it. No, thank you.
- [11:53]
B
For you, I want a little bit more. Well, a little bit more because I think this is a really good opportunity for your executive teams, okay? Which is. Hey, look, the bad guys are not omnipotent, okay? They're not all geniuses and they write the best possible code. There are people out there that are using vibe coding as citizen attackers. Like, we have citizen developers that are using this technology to try to make money, okay? And they look foolish in doing it because we're making fun of it. But the consequence to the business is actually worse because it's unrecoverable when. When they are successful, so they don't know what they're doing and they're actually creating more damage than the professionals. Okay?
- [12:48]
A
Yeah, yeah. They're not even competent at running their illicit business. That's a really interesting way to think about that.
- [12:56]
B
Okay. Yeah. Think about citizen developers in our lines of business that are building business applications that are unmaintainable, unsustainable, and unsecurable, but they sell more jeans, okay? So we have an equivalent problem that we've been talking about to our executive teams, but look at the consequence, right? Which was the business application functionally can also create a problem like allowing triple coupon codes on a checkout. And all of a sudden you're selling $100 jeans for a dollar. Okay? So the bad guys have the same problem, but for the bad guys, they may look foolish, but they actually hurt us as well. So it's a lose lose for the business people, for citizen developers and citizen malicious hackers.
- [13:51]
A
I do have to give props to Kevin Farrell in our chat for describing Sakari ransomware as actually brickware. Usually that's what I call my Linux machine after I try to update grub, but that's just me. Before we move on to our next story here, I have to spend a few moments and thank our sponsor for today. And that is strike 48. Security teams are stretched. Attack surfaces and threat volumes keep growing. Meanwhile, SOC budgets stay flat and glorified chatbots with hallucination problems aren't helping. Strike 48 is different agents scale independently running investigations across your logs, while your team can concentrate on the highest priority tasks that require human judgment and decision making. Try it today@strike48.com security. That's S T R I K E48.com security. All right, let's jump into the deep dives here. First up here. Judge dismisses Virginia Flock camera case. A federal judge upheld Norfolk, Virginia's use of 176 flock automated license plate reader cameras, rejecting claims that they amounted to unconstitutional warrantless surveillance. The court ruled the network is too sparse to reveal a whole picture of someone's movements, contrasting it with mobile phone tracking and aerial surveillance cases. The Institute for Justice, which brought the suit, plans to appeal as other cities end flock contracts over privacy concerns. It's an interesting interpretation of surveillance here. Nick, I'm going to start with you. Do you think it bodes badly for personal and corporate security? Do you agree that this doesn't give the whole picture to constitute warrantless surveillance? Where is your mind at with this?
- [15:30]
C
Yeah, I think this is absolutely a huge one. I hate this. For the record. Oh, Lord, do I hate this. This ruling is narrow. It basically it's not a blanket approval of mass surveillance, though. So like a federal judge, as you mentioned, basically is upholding the use of flock cameras because basically they looked at it and said it's too limited to construct full movement for a person. Right. If I just had flock and I'm walking through north Norfolk, you know, okay, great. You've got aspects of me but you don't have basically end to end. Right. But that hinges. That decision hinges on scale and aggregation because the court explicitly distinguished sparse fixed cameras from continuous tracking methods like mobile phone location, aerial surveillance, all of this. That implication, I think is deeply disturbing to me because not unconstitutional, quote, unquote, is not equal low risk in a scenario like this. Right. Camera density is going to increase. We know this flock is making tons of money, not to mention everything else. So that means data retention is going to expand. These systems are going to get integrated with other data sets. We've already seen that happening in the last year. So I think the judge totally missed the mark. And the long term of this is that aggregation, I think, is the real risk multiplier here.
- [16:48]
B
Right.
- [16:49]
C
Even just benign point in time, data can become sensitive when it's combined with other locations. Times third party sharing in 2018 basically. Ring doorbell. You know, Amazon ring went wide with like getting basically embedded in police. We've seen warrantless releases between ring and Nest and some of the others of this kind of data. So a flock's not picking me up. They've got other ways to get this and track me. Like, we can't look at these things in silos, you know. And I think that's one of the Big, big issues that we have here. And the Institute for Justice is planning to appeal this. Other municipalities are starting to end contracts over privacy concerns. You can look at Baltimore Francisco and like some of the others, but, but this is not settled and it shouldn't be. You know, and I want to be. This one just burns me because, like, I want to be very clear. I've been a privacy wonk for, for years. You know, in my position, I think this ruling does not green light broad surveillance. It is rewarding restraint. But the more comprehensive the picture that we, the privacy wonks can build, I think the higher legal and reputational risks these things are going to have. And I think that gives us a better chance to thwart these things long term, as long as we have a fair legal system to adjudicate it. It's a huge problem that we have. And we know if the government comes knocking, most of us are going to turn over that footage. It's of deep concern. I do not like this.
- [18:16]
A
And I would say the ability to correlate when you have different data sets is now even easier, as we've seen with, you know, when we're talking about LLMs and stuff like that, where you could feed that in theoretically and kind of build that. But Steve, I want to get your, I want to get your take on this. Like, how are you reading the tea leaves here with this, like Nick said. Nick. And I think that's a really accurate statement. You know, a fairly narrow ruling here that doesn't really set precedent too much going forward at least gives a minimum viable unconstitutionality barrier here. But Steve, what are your thoughts on this?
- [18:51]
B
This is an awesome two beer conversation because there's no right and there's no wrong. And why I say that being honest is greater good. The conversation here is what is the greater good? Is the greater good privacy at all cost, that we need to cease and desist all of this because the ability to aggregate, to use it for wrong, malicious with whatever your definition of malicious is, is the overriding concern. So therefore the greater good is we've got to stop this, okay? Whereas the greater good to appreciate the fact that I'm doing license plate scanning because I'm looking for right, some child was abducted in a 67 Honda, okay? And the camera scanning is now looking for that 67 Honda to save that poor 3 year old that just got abducted. Okay? So which is the better greater good? The ability to save the child or the ability to basically know that my data is protected and the government can't use it? Maliciously. I don't have a good answer because both are valid. Okay. The key is we don't trust the government with our data, rightly or wrongly. And so therefore this ability to be able to say you can't aggregate it. I understand. But in the independent cases, the use cases that are driving cameras, or our ability to aggregate data to do something for the greater good. That's what I said. What is the better greater good here for us to understand where the right middle ground is?
- [20:44]
A
Well, and I think when you have those conversations, I think where the frustration is for a lot of people is that there is nothing explicit being said in terms of consciously making that trade off. Right. There's a lot of implicit, and this comes up all the time when we talk about encryption security and all that kind of stuff. And we're talking about the exact same arguments apply. And I think there is, if we had more, we had the ability to sit down with a couple beers or drinks of choice, tomato juice if you want, whatever floats your boat, and consciously make those decisions. And now I realize in a democratic society, everyone kind of has to form some sort of agreement on it, but we get into problems when we don't specifically say it and we just assume it's binary. Right. It's total. Total this or total that. I, I, and I appreciate Steve saying not knowing where the balance is and kind of laying out the groundwork with it. You definitely have some fans in the chat with, kind of.
- [21:48]
C
Right.
- [21:48]
A
With where you're at with that.
- [21:50]
C
Well, if, if I can, if I can just add to that too, because I, I think, I think Steve's argument is very valid. It's one I hear quite a bit, especially when I'm debating privacy issues. I will say this, though. Nothing is absolute, and that was not my implication on this. What I am speaking to is we have a consistency. And by we, I mean, you know, those that have been historically and traditionally democratic in nature in terms of governance systems. So think Europe, United States, Canada, Australia, New Zealand, et cetera, where we have this desire in some way, shape or form to see a problem that statistically speaking isn't massive and then create laws to the lowest common denominator. I like to say that, that cybersecurity is agnostic to politics, but we're not immune from it. And this is exactly one of those cases where, to Steve's point, one, if you were able to abduct a child in a 67 Honda and it's running, congratulations. Because the 67 Honda is like, I don't think it would be running anyway.
- [22:48]
A
You know, this time of year you probably could run it.
- [22:50]
B
Yeah.
- [22:51]
A
Getting far in that car.
- [22:52]
C
Just, Just, you know, make something a little more reliable and new. Not that I'm advocating abducting children. The point is. But the point is, is that, is.
- [23:03]
A
That.
- [23:06]
C
The point is, is that we are making these laws of lowest common denominators when the vast majority of us are not criminals of any stripe. We're not robbing people, we're not breaking into things, we're not abducting kids or anybody. And all of that. Yet we have this reactionary need within government to create things that speak to that, that potentially impinge all of us. And to the encryption point, and why Apple in 2016 encryption over the San Bernardino shooting incident, I wrote articles on that one is because they knew that to capture data that goes stale from one person should not impact tens of millions of iPhone users. These are the kinds of trade offs that we have to understand and we have to make. But when you have a government, for example, any government that doesn't have good privacy laws, that have teeth in to ensure the government is separated from the data without a clear judicial process, something that has been breaking down in the last seven to 10 years. And again, I have tons of notes on this, you know, and we have to balance that with what the government is doing. And the government loves to overreach. Every major government says we should have a backdoor in encryption. Again, lowest common denominator. And I think that's something that we have to. We have to argue for, you know, and it's not that I want to see the kid kidnapped or have that person found, but we've also had law enforcement and government for decades that have not had advanced technology, you know, and we've always had this cat and mouse. And overarchingly the law enforcement has been successful when they need to be. So. So I just, I just don't see it as yet another erosion of a right over something that is statistically insignificant compared to the greater population, although very public. So we have, you know, we have.
- [25:02]
A
To get moving into our next. Right, but are both of our. We have some people in our chat that are both kind of in agreement here. Proper oversight definitely needed, in my amateur opinion. CCL and Kevin Farrell both saying that. And I think that if we want to talk about a way to have a middle ground, a way to have kind of a mediation right between two absolutes, I think that's, that's probably a good foundation for that next story Here Microsoft to disable NTLM by default in future Windows releases. New technology LAN Manager, despite what it says in the tin, is a 30 year old challenge response authentication protocol that was introduced in 1993 with Windows NT 3.1. It has now been superseded by Kerberos, which remains the default protocol for domain connected devices running Windows 2000 or later. NTLM is still used as a fallback authentication, although it has been widely exploited in ntlm. With the retirement of NTLM is part of Microsoft's push toward passwordless phishing resistant authentication methods and will occur in phases, as is the Microsoft tradition throughout 2026. This story shines a light on legacy systems and Steve kind of gets to what we were talking about earlier on with some of the vibe code or with some of the encryption stuff. Excuse me, and the perpetual problem of the discarded and therefore unpatched vulnerabilities that can exist in a network because people have forgotten about them. I'm curious though, Steve, how does this, how can we maybe draw some lessons here about dealing with legacy technology kind of going whether it's NTLM or picture poison here in this case, Steve.
- [26:37]
B
So given the number of systems out there still running Windows 95, it's amazing how long new technology can actually be new.
- [26:46]
A
It's true. Wow, love that.
- [26:52]
B
Just a perspective there, which was that is still new technology. I got to hand it to Microsoft, not bad. From a reality perspective, legacy is going to have a new conversation. It's going to be disconnected devices is the newer way we're talking about it. Disconnected or disassociated devices, of which legacy is just one facet of that. Because when we talked about citizen developers, look at all these applications that are not going to be connected but exist. So we're what's old is new again and we're starting to go there from the perspective of turning it off. Absolutely. Because if you have some of those apps, there comes a point where If I, after 25 years, can I get my company to retire that app because it's over an OT and it's untouchable. Thank you Microsoft for forcing the issue. Because now I can just basically, oh, Microsoft turned it off, I got no choice. So I can force the business finally to stop simply saying I don't have the money to spend because I get to blame Microsoft in order to improve our posture.
- [28:05]
A
I like the idea that Microsoft becoming once again comfortable with being the bad, with being the bad guy for people. That warms my Linux hippie heart. Nick, what did the story make you think? Like when it comes to legacy tech, ntlm, I said it right all these times. I'm not going to say it again because I'm going to get it wrong again. What are you thinking here?
- [28:24]
C
So my very first thought when I was reading about this one was, well, it's about damn time. You know, I mean, honestly, I mean, if you think about it, NTLM relay and pass the hash, super common attacks, you know, fallback authentication just becomes part of the entry point at that point. So, yeah, I'm all for this. I'm not going to lie. It's going to be tough for it and cybersecurity alike. To Steve's point about having that conversation with leadership about saying, hey, it's turned off by default, we need to kill this. You know, there's going to be executives out there that say, well, we'll just stay on Windows 2003, you know, indefinitely because we don't have to deal with all of this. So I think one of the big things is that, you know, to dovetail kind of what Steve was saying is I think this change forces visibility and into the technical debt side of any organization. Right? And we try and hide technical debt, you know, whether it's, you know, isolating networks of legacy systems. I mean, think about manufacturing, the amount of ot. I mean, I've seen stuff that came out when the Macarena was popular and these guys are not going to get rid of it until it breaks, you know, so. So I think it's going to be, I think it's going to have a good conversation, you know, and the new baseline of, you know, Kerberos and passwordless, you know, and that's not new, but, but that's becoming the baseline now. And I think that is super important. So if I were advising a ciso, I'd basically say, like, look, start inventorying your assets that require ntlm. You know, you're going to have legacy apps, I mean, printers, scanners, like all this kind of stuff, you know, that that's going to need this, and then start to plan for replacement or compensating controls in some way. But I think it's, I think overarchingly, it's a good thing. My only fear here, and this is a slightly different animal, obviously, is that Microsoft was really huge in pushing tls. One point, like, you know, oh, we're going to cut off 1.2, and we're still on 1.2, like after two, three years. So I'm hoping they don't backtrack this, even though I know it's going to be a pain point. You know, it's just we have to. And new technology has to get newer, right?
- [30:24]
A
I mean, Microsoft will still let you run SIFS if you really want to be a sadist. So, like, there's, there's, you know, there's, there's no end to. People who want to do damage to themselves can do damage.
- [30:36]
B
Now. I have to finish here. I know we're. But I have to be snarky about this going. Here's what they should have said. Microsoft to disable NTLM by default in future Windows releases to secure against quantum attacks.
- [30:49]
C
There you go.
- [30:50]
B
Why didn't they just do that?
- [30:51]
C
There you go.
- [30:52]
B
Okay. Because now we're right and everybody's like, oh, it's a quantum problem. And so therefore it's a good thing. They should have just done that. There you go.
- [31:01]
A
Well, wait a minute, wait a minute. Are you saying you can get anything greenlit? If you call it agentic quantum resistant resilience cloud, Nate. Like, is that, is that the secret buzzword sauce right now?
- [31:13]
B
Yeah.
- [31:13]
C
We should all just start a business called Quantum AI Solutions. And yes, just get funding.
- [31:19]
A
You know what? We actually did just get an angel round. Actually, now that I just checked, we actually are so, you know, I don't make the rules here it is. It is what it is.
- [31:31]
B
Throw them under the corner bus, we're good to go.
- [31:35]
A
Speaking of buzzy things here, this is a late breaking story, but we have to make sure we cover this today. And it's that Moltbot, it's just taken over the world at this point. A few days ago, Matt Schlicht Vibe coded a Reddit for AI agents called Multbook. We're all fascinated by it. Humans are view only on this thing. The bots are talking to themselves. In this case. This is all part of a wider conversation around this thing that was called claudebot, that is now Multbot, which is now openclaw. Basically, it's an AI agent that can have full autonomy basically over your system and kind of can do whatever you want it to do, including write its own apps for it to use to do the thing you ask to. That's a whole wider conversation. But when it comes to multiple books, security researchers, several of them, including those from Wiz and 404 Media, found a security exploit in its production database that exposed full read and write access. They exposed account owners, auth tokens, et cetera, and showed there were no controls in place to actually prevent humans from being on the service. They Found like a huge ratio. I think it was like 88 to. Or like 188 to 1. Account owners to bots, that kind of thing. Or bots to account owners. So I just got to ask Steve, this has just been eating up the world here. Why are we so fascinated by this? And, you know, is. Are we. Is there any other lessons from the. From the vibe coding thing of you can stand this thing up? That's very interesting. Very quickly, you may not be prepared for. Had that many eyeballs on it.
- [33:06]
B
So, you know, there's a phrase that says, don't run with knives. Okay.
- [33:14]
A
Or scissors. Or scissors.
- [33:15]
B
Or scissors. Okay. Now, as an adult, it makes perfect sense to say, don't run with scissors. Okay. Well, AI, if you think about it, depending upon who you are, is either at the age of a child or maybe the age of a teenager. So if it's bad to run with scissors as an adult, okay. And now we're giving it to AI that operates at a child or a teenager level. Are we not surprised that the outcome. Okay. Can be bad? Now, I actually appreciate what they did, and the security teams, of course, are jumping all over this because they want to sell security products. But I go as an exercise. This is why AI adoption, as amazing as it could be, requires us to take a step back and really kind of put the guardrails in. That simply says one, it's not a good idea to do so don't turn right the insane asylum loose here by basically having all the chatbots talk to each other because bad things are going to happen. But I think it's a really good opportunity for the industry to understand kind of where the state of the art is. Let us have a good laugh about it. Let the security teams be able to go, oh, my God. See, this is why you need to buy us. But it's a good reminder of kind of where the current state of capability is that we want to continue to look at. How do we mature the technology to get to a good outcome?
- [34:58]
A
Nick, jump in here. Where's your. What did this make you think of?
- [35:02]
C
I mean, do you know how moody teenagers are?
- [35:05]
B
There you go. See?
- [35:06]
C
And then you let them talk and coordinate against their parents. Like, this is how Skynet starts, man, I'm telling you. But honestly, I think the approach is kind of novel, and I think it's interesting to see. I mean, when I was quickly putting notes together on this, the first thing I thought of is that they're trying to create a thinking environment for artificial intelligence. Right. As opposed to just going to the chat, GPTs and Geminis of the world to ask a question, get an answer. They're trying to build something, I think bigger. And I think this would be good for research, analytics, all that kind of stu. But, you know, I just don't know where this is going to go. You know, I mean, I think academically it's an interesting and novel concept to do something like this because I think if I'm looking at this, this is more the journey than it is the end. Right. How did we get here, I think is just as important as the output of what this is ultimately going to look like. So if anything, this could be a maturity thing. And I think it also, you know, again, it's just, if I'm thinking this off the top of my head, I think it's where AI tools are going to be going. I think this is where they're heading. Right. You know, and I think it's going to go beyond chat only, you know, interactions. It's going to be a lot more persistent, you know, so I think it's going to, I think it is going to impact the future. I think something like this is probably the first iteration or one of the first iterations, but I don't think it's going to be the last. You know, I think it's, I think that's where we're heading here. Auditability, I think is going to be beyond important.
- [36:39]
A
But yeah, and that's, you know, in so many ways. Again, this is all moving so fast and we're all trying to keep ahead of it. This does very much remind me of where we were with AI image generation in 2024, where it was impossible not to have somebody with six fingers and stuff like that. Right. And then in 2025, if you told it, make sure it has this person has five fingers on each hand, it would get it right. But you needed to be smart enough to prompt it. And now it seems like, yes, you can still, it's still easy to ID AI slot, but it's, it's not as easy as it was, right?
- [37:07]
C
No, not at all.
- [37:08]
A
And so the fact that we're kind of seeing this, you know, this is like the third time, like third week in a row we've talked about, when we're talking about Sakari, we're talking about Void Link, we're talking about multiple, you know. Now at this point, I actually think the more interesting thing is the whole open claw project of, you know, where you have, have kind of this, this bot, that kind of has license to kind of go do whatever it wants, write that software that it needs to write to be able to do the task without you having to specify that. Like that, to me, is much more interesting and opens up a lot more questions of you're putting a. That is a heck of a point of failure. Right. With your weird telegram bot that you're talking to on your Mac mini that's sitting in your server. So there are so many. I mean, I think that's why this resonates with so many people. If you want this to be Skynet 1.0, this can be Skynet 1.0. If you want this to be the future of autonomous agents that are writing micro software that's specific to your organization, that's here too. Right. If you want to talk about the perils of vibe coding, that's here too. There's so many different angles in this that I find fascinating.
- [38:08]
C
Yeah, well, think about it. The big issue, if I'm just thinking about it overarchingly, is also model drift. These things can get completely off the rails at some point in time as well, you know, so then governance then becomes an issue of that as well. You know, I don't think, like, if I'm again, really thinking this through, I don't think Molt book is. It's not important because of its scale. I think it's because it's previewing, I think where AI is going. It's shifting basically AI from like a helper to kind of like a structured thinking system, I guess, is the best thing that I can come up with on that. Okay, So I mean, and to your point, a structured thinking system is. Exactly. I can do this, I can do that, you know, or I can pivot. And I don't necessarily know how good that is or how bad, but, you know, it's something that.
- [39:00]
B
To watch.
- [39:01]
C
It's interesting. Yeah.
- [39:03]
A
Oh, real quick, Steve, Let me just give some shout out to Schmooz in our chat. He said, before we empower artificial intelligence, why don't we do something about human stupidity? I think that was Socrates's argument to not write. Like, to not have a writing system. So that argument is eternal, but also very relevant. Steve, I'll let you get the last word here.
- [39:21]
B
Yep. And this is it. Which was. We've had all the A models for the last couple years, and we talk about red teaming, which was, how do we build a model? Okay. That that represents a synthetic identity. And we test it to make sure that we understand how it's going to Operate. Okay. To turn it loose. Well, human stupidity is the same way. How do I train a human so that I can trust it to operate in a standard way? Okay, now, what I didn't say is how can I do this so that I can trust the output? Okay. Because we want to trust it. We can't. And the way we evaluate models is not on is it trustable? Is and not is it telling the truth? But we talk about integrity, we talk about rightness. We have these six characteristics that we have defined as to how we train a model without actually saying, is it telling the truth or can I trust it? And that is state of the art. And this is an example of, well, we're just going to trust it and we say it's telling the truth to see how quickly it turns into an unsupportable outcome. Well, it's just state of the art. And this is our ability to create it, our ability to test it, and our ability to trust it. Just as I said, I think it's a great exercise now of what the state of the art is, and so therefore a valuable exercise. And I'm glad everybody's really taking a look at it because I think it's a timely conversation.
- [41:01]
C
Yeah. I mean, hey, if it improves my AI girlfriend, I mean, that's a win.
- [41:04]
A
So it's got a big data set now to try now. All right, well, you're getting upgraded.
- [41:11]
C
Sorry.
- [41:13]
A
Before we get out of here, besides what not to say to close out a podcast, is there any piece of advice you would like to share with our audience? Nick, I'm going to start with you. Where do you want to leave our audience with some words of wisdom, perhaps?
- [41:25]
C
Yeah, yeah. I think just with all those stories today, I mean, I think just overarchingly take a good hard look at where the world is going and plan accordingly. Understand the risk of everything that we've just talked about and quantify that to the best of your ability. And prepare for post quantum. Prepare for all of these other things, whether Steve yawns about them or not. I think that's my advice for everybody today.
- [41:54]
A
Steve, what about you? What are you going to leave our audience with?
- [41:57]
B
So, you know, false people, Most people know me as being the voice of reason. Okay. Take a look. Okay. What have we got to do to survive another day? What's the voice of reason? Where are we going? Let's understand this. And that's what today, I think, really was. Why I really enjoyed this conversation amongst us is evaluating. Here's the one thing I Leave you, which was audio and video. Social engineering attacks. Okay. Are escalating now. And so take a look at Microsoft Teams and others and see how the bad guys are now leveraging that capability. Okay. And what you're able to do to try to re establish a defensive perimeter there. I had this conversation today with the ciso and it was very frustrating to be able to see how little capability we have to allow our humans to be able to defend themselves or for us to be able to know it. And so my word of warning is on a practical level, those types of attacks are definitely on the forefront now. So spend a little time with your organizations and figure out what you're doing before you get hurt. That's it.
- [43:13]
A
Well, thank you so much to both of you for being on the show. Absolutely fantastic. Steve Zaluski, the co host of Defense in Depth here on the CISO series, where can. Can people see you out and about coming up here in the next couple months here?
- [43:28]
B
Yeah, really excited. RSA is coming in San Francisco. So to everybody out there, I go, I'm here in San Francisco, so I kind of like him, the host. I think of you as, hey, you're coming to my house. So I'm looking forward to seeing everybody. Love to be able to chat with folks. Right. See what else is going out there. So if you're coming to San Francisco and rsa, let me know. Happy to make some time, catch up.
- [43:54]
A
And before you say hi to Steve, you also have to go to our live CISO series podcast recording at BSides SF cisoseries.com events for more information. But Nick Espinoza, host of the Deep Dive Radio show, I know you've got some stuff going on. Where are you going to be? Out and about in the meatspace coming up.
- [44:11]
B
Yeah.
- [44:11]
C
Yeah. So I'm actually speaking at the world's largest construction conference called Conexpo in Las Vegas on. On March 7th. So if you want to see me basically on stage for an hour yelling at everybody, I'm more than happy to see you there. Should be a fun time.
- [44:27]
B
No, no. He's gonna be trying to drive the bulldozer outside. That's where you're gonna find Nick. He ain't interested in talking. Yeah, he wants to play with the.
- [44:37]
C
There's a lot of fun toys with these things.
- [44:38]
A
There really is technically ot so we could tie it together around here somehow.
- [44:43]
B
Figured out.
- [44:44]
A
It all works out. It all works out. We'll have links to both of your LinkedIn profiles in our show notes. So if you want to follow them and I highly recommend you do. You will have the resources to do so at your fingertips. Thanks also to our sponsor for today, Strike 48 Agentix Security. Start moving at machine speed. Remember, if you missed the live show, you can send us some feedback@feedbackisoseries.com we read everything, we respond and we have a good time. Remember to join us next Monday at 4pm Eastern for another edition of the Department of Know. To register for the live show on YouTube, head on over to cisoseries.com click on that events page like I already mentioned. You can find all the information there. It is awesome. Thank you so much for joining your Monday standup. We have. I hope you're having a great week lined up. You're staying warm, you're staying secure. For myself, for our glorious producer Steve Apprentice, for the big boss man David Spark and everyone else here with the CISO series, here's wishing you and yours to have a Super sparkly day. You.
- [45:44]
C
Cybersecurity headlines are available every weekday.
- [45:47]
A
Head to csoseries.com for the full stories behind the headlines.