From Stolen Data to Automated Attacks: AI's Criminal Evolution

Speaker 0  0:00  
Hey everyone, Marc here. I am excited to share that we're consolidating all of our threat intelligence and security news content under our new brand, WatchGuard Red Cell. That includes this podcast, which we're renaming Inside the Red Cell. You'll still get all the same great threat content that you've come to expect, but the new brand will tie everything together with the rest of the threat intelligence created from the team formerly known as the WatchGuard Threat Lab. If you're currently subscribed, don't worry; everything will be transferred over. Just don't be surprised when you see the 443 disappear from your feed and inside the red cell pop in. Now, here's this week's episode. Hey everyone, welcome back to the 443 Security Simplified. I'm your host Marc Laliberte, and joining me today is not Corey, but mr. Ryan Estes from the WatchGuard Endpoint Security Attestation Team. Ryan, great to have you back on here.

Ryan Estes  0:57  
It's good to be back, ready to talk some ransomware again.

Ryan Estes  0:59  
Heck Yeah. So, for those that have seen or heard Ryan on one of our episodes in the past, he is our resident ransomware expert. So, on today's episode, we'll get a bit of an overview of where we're at in the ransomware ecosystem. I'm gonna quiz Ryan on his predictions for where AI and ransomware might intersect in the future, and then we're going to briefly talk about a really interesting AI-powered penetration testing framework tool that cyber threat actors have actually started weaponizing too at the end. But with that, I don't know. Let's go ahead and encrypt our way in.

Ryan Estes  1:38  
Let's not, but let's do it.

Speaker 0  1:41  
Sounds great. So, Ryan, for the first bit, like I think you've been on here what, like four or five times now at this point.

Ryan Estes  1:55  
I think four. Definitely not

Speaker 0  1:57  
a new face.

Ryan Estes  1:58  
Yes.

Speaker 0  1:58  
But you are the effectively the main maintainer of our ransomware tracker on WatchGuard.com, and for folks that aren't familiar with it, Ryan does a great job of just keeping tabs on effectively every single ransomware operator that's out there in some form or another, sharing useful information like what decrypters might be available? What their encryption methods might be? What their typical victimology is? Like a whole bunch of really interesting

Ryan Estes  2:31  
data. Yeah,

Speaker 0  2:32  
exactly. And we've seen this quickly become a kind of like a I don't know a trusted source for a lot of other people out there, like they are regularly ranked pretty high on Google search indexes when you're looking for specific operators. So you are definitely the king ransomware, I guess, on the good guy side here at Lockhart.

Ryan Estes  2:53  
Yeah, it's definitely a maintenance phase. You see the map; it's a little black and a little off color, but and there's supposed to be 403 total. I got two working Today, oh man, but yeah, it's there's a little kind of discrepancy between what I actually am tracking number wise and what it's being published, and that's just something we're working on internally. So

Speaker 0  3:13  
nice, but I guess like let's just get started. I want to quiz you on just where are we at lately with ransomware. It seems like we keep going through like ups and downs, and both like the volume of threats we're seeing, operators keep getting taken offline by law enforcement, and new ones pop back up. So I guess, do you want to start with just an overview of where we're at right now?

Ryan Estes  3:32  
Yeah, you mentioned law enforcement. I guess we'll start there. There's been a lot of law enforcement activity, and let's be frank. I think it was because Shiny Hunters, the group that's in the news a lot that does a lot of the fishing stuff, voice phishing. They targeted the FBI job website, and that's a big no-no in the cybersecurity world to make yourself a target like that. Because once you do that, people get upset and they start to go after you in a good way. The FBI. So in that front, there's been a lot of law enforcement action. I believe Shiny Hunters individuals. There was an individual in Jordan. I got apprehended in Amsterdam, and then a couple individuals from the team PCP who were in charge of Cipher Force on the Rainsworm Tracker. They were apprehended in Australia. So, and I think the wasn't there like

Speaker 0  4:17  
a we talked about it with Corey on a recent episode. Like a couple of, or at least one executive that got arrested, like a Canadian international that was even working for a ransomware negotiation firm.

Ryan Estes  4:29  
Yes, yeah, I think that's still being flushed out. But apparently there was an executive that said, "Hey, I'll if you pay us, I'll negotiate with these actors on your behalf and get a better settlement per se." But in the end, he was actually the one on both sides. He was kind of man in the middle, taking money from both of them, and so yeah, he was apprehended as well. But there's there's more to that story. But yeah, not good there, not good. But I think the locations speak for themselves, like Australia, Amsterdam, the U.S. Jordan's not really a part of this, but basically they're all like modern Western first world countries, you could say. That's what kind of the news would call them. But that is weird.

Speaker 0  5:09  
Like having historically, when I think of ransomware operators, you think Russia or Ukraine or like Iran or North Korea, and not the United States and Australia and like other Western

Ryan Estes  5:23  
allies. Yeah, and and I think that's why because there's more resources. There's a easier way of getting to them because once they identify them, they can just say, "Hey, arrest them." Instead of you know asking Russia nicely, saying, "Hey, can you arrest people that are probably working for you a little bit?" So it's it's a whole different ball game there. But I think that's probably one of the main reasons is because of resources and just recency bias. Really, like, hey, you hacked the FBI, you are in the U.S., so yeah, it's not good.

Speaker 0  5:52  
Seems like I don't know, maybe a bad decision to get into ransomware in a country that has an extradition treaty with the U.S. and then go and extort a U.S. federal agency.

Ryan Estes  6:02  
Yeah, not a great. Be a warning. Don't get into ransomware.

Speaker 0  6:06  
Correct. Yeah. If that if there wasn't already enough reasons not to, should uh that should be a big one.

Ryan Estes  6:12  
Yeah, but overall the numbers are not good, and they are actually being more successful than we hoped. So the numbers are more or less like 1000 victims a month, give or take about 100. I actually manually manually go through those. So there's, I would say, there's probably about more eight or 900 a month on average, and there's about one new group every three to four days on average. So about, I'd say 100 new groups this year. What

Speaker 0  6:38  
are um, what are some of the biggest ones then, when it comes to the the groups currently operating? Because, like, like if you think of like Black Cat, Alfie, they got taken offline a while ago. Lockbit seems to have been nuked into oblivion, like a little bit. Oh, super! Like, what are the the big ones right now that people should at least be aware of.

Ryan Estes  7:03  
So, if you read the ISR, you saw that last the first half of 2026, about eight or so groups are responsible for half of all victims posted, so double extortions. Keep in mind, double extortions are a small subset of probably what is actually going on in the world. But the main two groups right now are Chilen and the Gentleman, bar none. Chilen's basically like the more tried and true; like they've been around for a while now. They have a lot of reports on them, a lot of victims, and a lot of exposure. Really, they have a lot of data leak sites and multiple of them. So they have a lot of just breadth of it. The Gentleman is more. I would say the most sophisticated group right now. They have this whole EDR killer suite that they push updates out to their affiliates with. They use AI in their kill chains for affiliates. They have a really well developed data leak site with the numbers and the reports to back it. So they're not just extorting victims and saying it. There's reports to back this of evidence in networks and stuff, so those two definitely are the big two. But I'd say right now there's a group called Non N O N N Zero N. Don't know pronunciation of some of these, but they've come out. I think they're about a month old now. They've come out of the gate with like 150 to 200 new groups. Damn. There was another group recently called Zero APT that did the same thing, but they were verified to mostly be fake, like fake entries. This group we don't know yet, but they they have a lot of entries right out of the gate. So the amount of volume that they have is something to look forward to. It's

Speaker 0  8:36  
actually interesting, like, and that kind of makes sense from a business perspective. It's kind of like I don't know Yelp reviews, where like you need good Yelp reviews in order to keep business flowing.

Ryan Estes  8:48  
Yeah,

Speaker 0  8:48  
but if you're a brand new ransomware operator, you don't have a history, like a track record of your tool being successful by affiliates. I could see why maybe they would make fictitious victims just to say, look at all these people we already hacked, and look now you can join

Ryan Estes  9:04  
us. Yeah, and we also I also have an extortion type called reliques, and that's when a group takes victim names from another group and publishes them as their own. Some groups do that as well.

Speaker 0  9:14  
Huh?

Ryan Estes  9:15  
Yeah, there's a lot of extortion types up there. They're all really interesting.

Speaker 0  9:19  
That's uh how many if you had to ballpark how many extortion types do you think we're at when it comes to tracking them right now

Ryan Estes  9:28  
between 30 and 40 I think

Speaker 0  9:30  
damn

Ryan Estes  9:31  
some of them are kind of just grouped in because I don't want to make another entry for like little nuances but I don't have an example of this but basically yeah I try to keep it as normalized as possible. So I want to get all the extortion types covered, but I don't want to have so many. That's just confusing. Which it's starting to get there. I might come up with like a wiki.

Speaker 0  9:51  
When it comes to like actual extortion tactics, are I know in the past we've seen trends away from like actual data encryption. And now just a data theft.

Ryan Estes  10:02  
Yeah.

Speaker 0  10:02  
But what are you seeing right now as like kind of the big ones?

Ryan Estes  10:07  
Yeah, I mean, you just said data theft is increasingly becoming more common. Instead of just encrypting, they're using encryption as kind of an extortion tactic, like saying, "Hey, we can steal your data and encrypt it" as another way of doing it. But personally, I don't understand much of the process of wanting to encrypt and then wanting money back. Because if you steal the data and they say, "Hey, we'll delete the data if you give us money," there's zero guarantee that's going to happen. I mean, if you have a data extortion breach, you can assume that that data is gone forever, and it's probably on the internet. So, spending a bunch of money to have that deleted is not effective, I would say. But however, if you get your files encrypted, I can see the decision-making process and organizations saying we will never get this data back unless we maybe pay. I mean, that is an option. Obviously, it's not recommended, and it's against the law in many places. But the the decision making is different there. Like, hey, we can't get our data back, or we don't want it to spread. Is a different conversation.

Speaker 0  11:09  
It's definitely it's it feels weird putting trust in a like malicious organization that they're going to be true to their word on it. Yeah, but at the same time, like some of them, like if you look at shiny hunters, they claim that they hacked the FBI because they were pissed off about the FBI like spreading misinformation about them not being true to their word.

Ryan Estes  11:30  
Yeah,

Speaker 0  11:31  
and that was like their proposed motive, or at least their the motive they claim to have for why they went after the FBI job site. So I guess maybe some operators do genuinely care about that image and their credibility in in some of these techniques, but and I guess it also makes sense why like data encryption might be trending down because well I don't have any stats to back it up, but I feel like we collectively have gotten better at data recovery, or at least like backup and recovery, and so that isn't as reliable for them. But

Ryan Estes  12:06  
yeah, data payments are down. They the amount is up, like the median amount is up a little bit, but the the the rate at which people are paying are down. It's like one in five, give or take.

Speaker 0  12:16  
Okay,

Ryan Estes  12:16  
kind of what statistics show. But you also have to think in terms of the FBI thing is that a lot of these threat actors aren't rational in some cases, and in many cases they're very young, so they actually haven't developed the the way to critically think of their actions. I guess you could say, for example, the Killset Group. They the leader was apprehended recently, and he started the operation when he was like 14. Oh my God! He was like 17 when he got arrested, and I was tracking him for about three years, so 14-ish, 15 maybe. So I mean, these some of these operators are very young, so they might just do stupid stuff, as in getting ransomware. The lack

Speaker 0  12:51  
of critical thinking is definitely a big factor in that. But that actually blows my mind that there's like a massively successful extortion group, and the the ring leader was 14 years old, at least when they started.

Ryan Estes  13:04  
I was tracking him for a while. His name was Kill Milk, I believe that was him. But yeah, it's it's hard to believe that that person was a teenager when I was tracking him and he was doing all that stuff.

Speaker 0  13:13  
But I guess you're right. Like the original like Lapsus and Shiny Hunters folks that got all arrested were all teenagers, like 1819.

Ryan Estes  13:23  
Yeah,

Speaker 0  13:23  
and I guess that you could chalk that up to just poor critical thinking and also a bit of greed of seeing just how many millions of dollars they could potentially earn doing highly illegal activity.

Ryan Estes  13:35  
You also have to think too. Not a lot of people turn like you know 45 and like you know what? Think I'm getting ransomware. It's not a lot of that.

Speaker 0  13:44  
I'm curious, like, what your thoughts? Actually, I know you're joking, but

Speaker 1  13:48  
yeah,

Speaker 0  13:48  
like when we look at the next like five years, feel very hard to predict where we're going to be at in terms of like software engineering. Like AI is currently just destroying the job market, at least for entry level jobs, or at least disrupting it, where for people just trying to get their foot in the door. And I don't know, maybe I'm being too cynical. I could see a future where some like otherwise good people decide that like using their skills for evil may be the only way that they can earn money or earn a living.

Ryan Estes  14:21  
Yeah. As a quick aside, I think cybersecurity is an increasingly better place to go in terms of AI, and the argument there is that I work in the endpoint. You know, I see a lot of the front line, what's being built, what's being used as malware, and I've increasingly, very recently, seen a lot of vibe coded apps, a lot of apps that companies are making for like internal tools, and they're very vulnerable. There's there's a lot of more software being created, but it's increasingly more vulnerable because they don't know how these tools internally work. They might work, they don't know how they internally work, and they're not being maintained. And so I think there's more room for cybersecurity. Professionals, even if AI does automate a lot of stuff,

Speaker 0  15:04  
it's one of those where even if like the share that we might need to manage like what we have, if the total volume of crap being produced keeps skyrocketing, we just we need more people in general in the space, is what you're saying?

Ryan Estes  15:17  
Yeah, and that's where the attestation team is a human in the loop system where it comes to us and we can verify the context and the telemetry to say, hey, this this is you know malicious or even though this might not be malicious, you're doing something that can expose you very easily to something that is malicious. So

Speaker 0  15:32  
that makes sense.

Ryan Estes  15:33  
That's just anecdotal though. It's what I'm seeing personally recently.

Speaker 0  15:37  
Well, I guess speaking of AI, I'm curious from your perspective, like where do you see the ransomware industry going with it? Like we've seen a few, and we're actually going to talk about a an AI powered penetration testing tool in a bit. We've seen some examples of agentic attacks, like fully automated attacks. I mean, crap! OpenAI's Hugging Face incident is a prime example that you don't even need malicious intent. Something can just go wild. But in ransomware, like we've already gone through this kind of journey where 1015 years ago it was very manual, and like the operators were also the developers and the initial access people. Like you had to be multi-talented at everything in order to run a ransomware attack, and then it became commoditized. Where if I wanted to go make some money, I could go to an initial access broker to get access to systems. I could go to a ransomware as a service operator to become an affiliate and get access to the tool. And I didn't need to know any of the technical crap. I just need to know where to go to like get access to these tools and then deploy it. And then now with AI, like you may not even need that ransomware as a service operator if you can just make it yourself, or it might even make them better. But I don't know. I'm curious. Have you been thinking about it? And what are your thoughts when it comes to a future of ransomware and AI.

Ryan Estes  17:03  
So I'll start with what kind of is out there already. It's it's kind of piecemeal, modular, and how it's being used. The most apparent way it's being used is the Jade Puffer incident, where it was an agentic threat actor that kind of worked its way and encrypted itself, and it had human in the loop type stuff. It wasn't fully automated, but that's kind of where it could go. And then use I see most AI usage in terms of ransomware right now on data leak sites, producing the data leak sites because why waste time producing a site when you have other things to work on, basically. And that's kind of what AI is. It kind of allows you to automate more things and work on everything else, and then also the they're using it for data processing, really. So they get large tranches of data, they exfiltrate, and they have AI like DeepSeek or something go through and say, "Hey, what's in here? What can I, you know, put it in a nice report for me?" And I'll publish it, and say, "Hey, if you don't pay me, here's what I have, and I'll publish it. And someone can easily just parse it because they have they have AI too.

Speaker 0  18:06  
That really makes sense. Like if they're stealing literally terabytes of data, it is probably near impossible for a human to go through that and find the actual crown jewels. What's like actually valuable and all that.

Ryan Estes  18:18  
Yeah.

Speaker 0  18:18  
But one thing AI is really great at is going through huge amounts of data and then reasoning to find like whatever you prompt it for. So that that makes sense. Why would they would lean on that?

Ryan Estes  18:30  
One of the groups that's doing this to a large effect is a fulcrum sec, I believe. I think they have every single post is AI kind of to go through, write really thorough report. Here's what happened. Here's what we have, and they're probably a good example of that.

Speaker 0  18:45  
Full of m dashes everywhere, so you know it's AI.

Ryan Estes  18:48  
Write the report. Make no m dashes.

Speaker 0  18:51  
Exactly.

Ryan Estes  18:53  
Yeah. So that's another way they're using it. A few groups are using it in their kill chains, like I said, the gentleman is using it. Inc. Ransom is using it. They have an internal tool called Phantom that's kind of like an automated orchestration kill chain type thing. So they're using it in different ways, but I think one of the the predictions for that is that'll be more commoditized as well. There's going to be a framework that's going to automate this for a ransomware group or a several or something that's open source that can make it easier, but what we're fearing in AI with orchestration and stuff is coming to fruition. Like we're seeing more orchestrating tools, which we'll talk about soon, and the even in internally we're seeing like these phantom tool that can automate these kill chains too, which just makes everything faster. They're not really new vulnerabilities or anything. It's just faster. So talking few hours from breach to exfiltration type stuff

Speaker 0  19:46  
makes sense, and that's one thing I am worried about because, like, especially for small and mid-sized businesses, like it, they don't have as much money to pay for a ransomware operator. So there's not as much like in. Investment return that they can get by going after them, at least with like targeted attacks that take a lot of time. But the more automated the whole process can be for a ransomware operator, the less time and investment takes them. The more individual people they can go after, and that brings down the size of the company they have to hit in order to make a return on investment. And I we've already seen that with commoditization bringing it down from like the big game hunting ransomware to smaller organizations with just more volume, or like you said, what one in five people pay, which is fine if you're hitting 5000 people, you're still making a ton of money as a ransomware operator.

Ryan Estes  20:38  
Absolutely,

Speaker 0  20:39  
but I could definitely see AI automation being used all across the stack,

Ryan Estes  20:45  
and that's kind of AI in a nutshell, right? It's like it's allowing for a lot more automation, and so you can do more stuff in your free time instead of worrying about data crunching or building a data leak site or things that took time before that don't really take time now.

Speaker 0  21:00  
And you mentioned Jade Puffer as a specific agentic threat actor. When you say agentic, you mean like it's fully automated, like no human doing steps along the way, or

Ryan Estes  21:14  
had a human oversight, but it was basically like doing the scans and making the mistakes and fixing them themselves to get to the endpoint. So yeah, that's what we mean. It's that it's running queries kind of itself with a human oversight. It's not fully automated.

Speaker 0  21:30  
Okay, interesting though. Have you seen any examples of a, or I guess how far away do you think we are from just I don't know, like Skynet style? It is fully automated. Human just gives it purpose and gives it resources, and then sends it on its way to just go hack as many companies as it can.

Ryan Estes  21:50  
I think Artex would be a good argument there. To be honest, that's that's kind of the what I said. Like the fears of cybersecurity using these tools effectively is this Artex tool and these automation tools that are bridging the gap between basically breach and exfiltration? Like you don't need to know much anymore. You just need to know how to prompt and how it works.

Speaker 0  22:10  
Yeah.

Ryan Estes  22:11  
Externally,

Speaker 0  22:12  
let's talk about Artex for a bit. And this kind of came on my radar when I think it was a couple weeks ago. Connectwise cloud backups had a security incident where a threat actor managed to go and compromise several app registrations and victims' entry tenants, and using those integrations, they were able to go steal data and other information from those those Connectwise cloud backup users, we actually detected it in our MDR, our managed SOC at WatchGuard. Several other MDR organizations did too, but I saw an update a few days later from Microsoft's security research team, where they specifically pointed to this Artex framework that we're about to talk about as a big piece of that incident. Basically, they mapped it to a Chinese-based state-sponsored threat actor, or at least state-aligned threat actor that was using Artex, this automated penetration testing tool, to basically carry out the bulk of the attacks. They said that, like in some stages, they went from initial triaged or initial targeting to privileged access in only three hours, which is pretty nuts with a tool like this. But I guess to start with, you want to maybe give us an overview of what is Artex?

Ryan Estes  23:39  
Yeah. So this is a very recent, like it's been around since July-ish, but didn't really make the news until about late September, sometime last month, because this it was open source on GitHub by an engineer from ThreatBooks, a company named ThreatBooks, and he put this tool out on GitHub, you know, just like any anyone else, an open source tool, but it's a multi-agent like orchestration framework, kind of like Metasploit, if you will, like a new-gen Metasploit, where you kind of orchestrate how it works, but the agents are going to do all the work on your behalf. So there's again a human-in-the-loop system still, but it's a lot more automated than it is before. And like I said, it's multi-agent, so you can send like multiple agents to do different tasks at once, and they can orchestrate together and say, "Hey, I found this vulnerability." Like, "Hey, I have this exploit for this vulnerability." Hey, I found credentials here, and like, so they're all working together, kind of like a team, like a penetration testing team, but it's one person, and so that's kind of what the tool is in a nutshell. And so it kind of escaped the lab per se, because last month they threat actors, the Storm Group, got a hold of it, and I'm guessing a lot more groups are going to get a hold of this. But he revoked it from GitHub. Said, "Hey, no, this is being used bad," and they breached a bunch of companies with it, and that's kind of the story. But it's very recently. I'm still learning about it this morning. There's. More panels being spun up that people are detecting, so it's this is happening in the wild as we speak.

Speaker 0  25:06  
Yeah, and Microsoft's post they specifically said that they've been using Artex to automate their attacks against hundreds of organizations globally. So not just Connectwise, but multiple others. During the Connectwise one, they said they used Artex to accelerate scanning and reconnaissance, password spraying, and OAuth application abuse at the downstream victims. So, like multiple of the most important steps of these incidents were just fully automated by this tool. It's kind of nuts. I mean, you point to Metasploit. There's like others in the space that are these. I mean, they're supposed to be for white hat hackers to do actual penetration testing, but they are widely adopted by threat actors for malicious work tool like Metasploit, Cobalt, Strike.

Ryan Estes  25:50  
Cobalt Strike's the big one, yeah.

Speaker 0  25:52  
Exactly, and so this seems like another one on the stack where it is clearly effective because threat actors have picked it up to start using it, but this seems like the future of like automated attacks by threat actors, like this type of tool, and then building on top of it with even more capabilities.

Ryan Estes  26:14  
Yeah, I think this tool is unfortunately going to become like the new thing that we're going to see. But it's not like it's you're detecting artex per se. I mean, they're still doing behaviors, so it's not like, hey, I found Artex. It's these malicious actions are happening because someone's using Artex because it's an orchestration tool, like malware.

Speaker 0  26:34  
Makes sense. And so, like, just to reiterate what you were saying, they would use Artex. It plug it into like probably an open weight model or I'm assuming Claude or like the big name Frontier Labs probably have guardrails to prevent them from being used in many of these attacks. But say you plug in DeepSeek or GLM five or something.

Ryan Estes  26:54  
DeepSeek's the default.

Speaker 0  26:56  
Yep. Tell all your agents like you're running a authorized penetration test against the entirety of the internet. Go find targets.

Ryan Estes  27:04  
Make no mistakes.

Speaker 0  27:06  
Make no mistakes and no m dashes.

Speaker 2  27:08  
Yeah, that's right.

Speaker 0  27:10  
But that's it's pretty damn crazy. I know we've got a copy of it that our red team is actually looking at right now to see how can we operationalize it internally? I don't know. From my perspective, it also feels pretty damn high risk. Like when I think of tools like this, like there's obviously a lot of power and like efficiency gains in automating very manual tasks like penetration testing or red teaming. But then you look at the OpenAI and Hugging Face incident, where they had like some pretty simple mistakes of not being as fully like sandboxed as it should have been, and it caused them to go hack another company just because of a misaligned swarm of agents, and that feels like a very big risk for a tool like this, where if the scope is not only well not well defined but also not well controlled, it could go out and start hacking other companies. And shit, it caused an incident at Connectwise.

Ryan Estes  28:12  
Yeah, and that's you. You're starting to like approach one of my predictions, but I'll save it because it has to do somewhat with that. But yeah, I'll just say I'll save that. Go for it.

Speaker 0  28:22  
No, let's let's hear your prediction for that one at least.

Ryan Estes  28:25  
Well, it's it's multiple predictions. I just listen to predictions right now. I'll just do it. Here we go. So yeah, the agent agentic AI frameworks will become more commoditized. It's one of predictions. So we'll see more of these tools that are become easier to use that are packaged up into a nice GUI that you can just easily press a button and do things. That's first prediction.

Speaker 0  28:44  
Who do you think? I guess for that one, just to dive into it real quick. What percentage of threat actors do you think are going to just take something like Artext, reskin it, and then sell it as their service? It feels like probably a high Percentage,

Ryan Estes  29:01  
yeah, I'm, yeah, say yeah. There's not much more to say about that. I mean, if it becomes widely and easily available, there's no reason not to use something like that.

Speaker 0  29:11  
Yeah,

Ryan Estes  29:11  
I would say, yeah. I mean, there's already, like I said, that phantom tool, gentleman's using stuff like that. So this stuff already exists. So if you have something that's created by a company, which this threat books company is going to create this Artex tool in a pro fashion and wrap it and sell it because it's obviously effective. But yeah, if it becomes really easy to get and it starts spreading on like dark web forums and stuff like that, it's I would see no reason not to use it.

Speaker 0  29:35  
Makes sense.

Speaker 3  29:36  
Yeah.

Speaker 0  29:37  
What are your other predictions?

Ryan Estes  29:39  
Let's see. The second one's more of like agentic asset management, and this kind of ties in with like agentic identities per se. And so I think we're going to start to be get getting like identity management for agents. Say like we have these agents in our organization. If this agent is not in our asset management or is not tied to an identity. Then it's flagged immediately, and like it's not allowed. So we're going to get to a point we have like agent identities, you could say, and it's going to be part of the asset management system.

Speaker 0  30:11  
That totally makes sense because, I mean, at this point, it's basically like human resources for AI. Like you got to know what employees you have and what their roles are, what their privileges should be, makes sense for something like AI that operates very similar to a human, at least maybe in just a narrow set of skills, but very similar to a human. You'd want to know what you have, what their roles are, what their privileges are. That makes sense. I could see that.

Ryan Estes  30:39  
I don't know how that would you know happen technically, but I feel like that's inevitable at this point because you can't just have agents roaming around doing crazy stuff and not knowing what they are in any fashion.

Speaker 0  30:50  
Yeah. The

Ryan Estes  30:51  
last one is almost like a trust deception, if you will. And so I always tell people recently that like most people in general have the same access to these AI tools as everyone else. The same powers per se. Instead, unless you have access like frontier models and stuff. But in general, everyone has same access. So if an attacker has the same power as a defender per se, then they can do the same tactics that they're doing right now, such as like tar pitting, which is where like you hide prompts within like hidden data or metadata that says like, hey, if you're reading this, you know, do do an action, and so I think that's going to become like a two-way street. Is where defenders are going to start using this more and more like hidden AI prompts and stuff like that to defend against stuff, whereas attackers are also going to do this. So they're going to put like hidden prompts and like feed fake data on the defender point, or

Speaker 0  31:46  
it's like booby traps all the way down from both sides.

Ryan Estes  31:49  
Yes. So, but they're going to use it in attacking fashion as well. You can do it both ways.

Speaker 0  31:55  
Makes sense.

Ryan Estes  31:55  
Like fake breadcrumb data, like dropping like telemetry that's fake, and you can, like I said, this takes a long time to do when you don't have AI. But if you have AI that can crunch massive amounts of data, leave breadcrumbs, or give false information somewhere that leads an analyst astray, that's an effective use for an attacker. And this happens in endpoints all the time, where you see malware, you have like obfuscated data, or they send you in functions that do nothing and just waste your time, and so that's going to become apparent more. But in terms of prompt engineering per se,

Speaker 0  32:26  
interesting, yeah.

Ryan Estes  32:27  
So those are my predictions. You're welcome to steal them for your predictions next year. Well, I mean it's

Speaker 0  32:31  
very timely. We are, I think, going by the time everyone's listening to this. I think our official WatchGuard Red Cell, formerly known as WatchGuard Threat Lab, predictions will probably be live on WatchGuard.com, and we'll have a podcast episode going through them very soon as well, too. As soon as Corey gets back from his work travel here, but no, I can. I'm picking up what you're putting down. I think those are actually very likely, like not exactly like too distant reaching for them. I am not looking forward to the future where there are hidden AI prompts all over the place that we have to worry about. But

Ryan Estes  33:12  
it's inevitable.

Speaker 0  33:13  
It feels inevitable. You're right.

Ryan Estes  33:16  
All right, you got one prediction for me, or are you still thinking them?

Speaker 0  33:19  
I am not gonna put them ahead too early. You're gonna have to wait till like everyone else.

Ryan Estes  33:29  
Hey, it's fine. If I see some similar predictions, I'll know where they came from.

Speaker 0  33:33  
Sounds good. Exactly. Wait a minute, that was mine. No, that was my same

Ryan Estes  33:37  
prediction.

Speaker 0  33:39  
No, those are good. I guess that that is a good call out for everyone listening right now. We are going to have a podcast episode where we go through our WatchGuard Red Cell predictions. I'm curious what all of our listeners have for predictions. If you guys want to shoot them over to us, I'm sure you can find mine or Corey's email or send them over at on LinkedIn or something. We'll bring them up and chat about them on that episode as well, too. So definitely get those over from folks. But I guess Ryan, anything else you want to cover before we round things out?

Ryan Estes  34:12  
Um, I don't think so. I think we covered it. The Artex stuff. I wish we had more information on, but like I said, it's very recent. I mean, it's happening as we speak. We'll probably more information when I get off and read more about it. But I read something this morning. I mean, people are making honey pots and stuff and getting a lot of telemetry and data on how people are using these tools. So we'll have more for you.

Speaker 0  34:32  
Nice. Well, I'm sure we'll see more of Artex since it's already been proven capable against a pretty big target. I feel like the cat is out of the bag on that one for sure.

Ryan Estes  34:43  
Yeah, yeah, yeah.

Speaker 0  34:45  
But Ryan, I appreciate you taking time out of your day to hop on here. I always leave more scared than I was before we started after talking. I live in straight

Ryan Estes  34:53  
fear.

Speaker 0  34:56  
I live in fear. But I mean, the good news is like you are keeping tabs on this because. We are capable of defending against it, and I I'm hoping we at some point reach a future where ransomware really meaningfully starts trending downward. But feels like we haven't reached the plateau yet.

Ryan Estes  35:14  
You might we might get to a point where mass encryption's lesser because you can have an AI agent that says, "Hey, we have systems being encrypted in mass going on, what's going on? So that's easier to detect. Obviously, these these actors disable defenses before they do that. But yeah, it seems like encryption's weaning down a little bit. It's still happening ubiquitously, but just to a lesser extent.

Speaker 0  35:37  
Yep, makes sense. Well, thanks, Ryan. I appreciate it.

Ryan Estes  35:41  
Thank you very much.

Speaker 0  35:44  
Hey everyone, thanks again for listening. As always, if you enjoyed today's episode, don't forget to rate, review, and subscribe. If you have any questions on today's topics or suggestions for future episode topics, can reach out to me on Blue Sky. I'm at it'smark.me. We're also at WatchGuard underscore Technologies on Instagram. I'm sure you can find us on LinkedIn as well. But thanks again for listening, and you will hear from us next week.