37 minSeason 3, episode 19

Can AI Replace Security Teams? The Software Quality Debate

With Chris Romeo, Izar Tarandach, and Matt Coles

Secure software development

Is the cybersecurity industry facing a security problem or a software quality problem? In this episode, we’re tackling the controversial claim that AI advancements could make security teams obsolete—and uncover the deeper issues plaguing software development.

Episode title image: Can AI Replace Security Teams? The Software Quality Debate

Listen to the conversation

About this episode

Is the cybersecurity industry facing a security problem or a software quality problem? In this episode, we’re tackling the controversial claim that AI advancements could make security teams obsolete—and uncover the deeper issues plaguing software development. The conversation reveals an uncomfortable truth: software companies often transfer the risk of vulnerabilities to customers, creating a system where there's little incentive to invest in security by design. Can AI bridge this gap, or do we need fundamental changes in how we approach software development and regulation?

Article: Ex-CISA head thinks AI might fix code so fast we won't need security teams




Follow AI Security Table:

➜ Home: https://securitytable.ai/

➜ X: https://x.com/AISecurityTable

➜ LinkedIn: https://www.linkedin.com/company/ai-security-table/

➜ YouTube: https://www.youtube.com/@AISecurityTable

Transcript · 37 min conversation

0:11Chris Romeo Hey folks, welcome to another episode of the Security Table. This is Chris Romeo, joined by Isar Terrindosh and Matt Coles. And if you're watching this on video, you'll see that Isar has just returned from some form of conference season because he's fully decked out in T-shirt, uh, polar— I don't know what you call that thing— sweater, polar— I don't know, it's early in the morning where I live.

0:34Izar Tarandach So it has to be pointed out that, uh, not our sponsors yet, but, uh, they do have the best swag. Phoenix Security.

0:42Chris Romeo Yeah, there you go. I'm a fan.

0:44Izar Tarandach Black sweater for the win.

0:46Chris Romeo I'm a fan of those. And you got the collar turned up like the Fonz.

0:49Izar Tarandach So yeah, yeah, this one is to go like Risk? I don't know what risk is.

0:56Chris Romeo All you need is a pipe, you know, if you have a little pipe to, you know, kind of sit back and ponder.

1:01Izar Tarandach He's like, "Arrr." Arrr?

1:05Matt Coles Did you say arrr?

1:06Izar Tarandach Oh, we met Captain Morgan over there.

1:09Chris Romeo Whoa, whoa, whoa, hold on. The real Captain Morgan, the dude from the side of the bottle.

1:13Izar Tarandach The real Captain Morgan was there. Actually, the real Captain Morgan was everywhere there.

1:17Chris Romeo Is this like a metaphor?

1:20Izar Tarandach Nope. No.

1:20Chris Romeo Okay.

1:21Izar Tarandach Captain Morgan was actually there.

1:23Matt Coles He's not, he's not that subtle.

1:24Izar Tarandach Okay.

1:26Chris Romeo No. Matt, after, after working with Izar for 20 years, 20 plus years, I don't know. He's not that subtle. That's easy to tell.

1:37Izar Tarandach Not known for it.

1:38Chris Romeo How would you describe Izar Terran Dash, Matt?

1:42Matt Coles That's so good.

1:44Chris Romeo Such a good definition.

1:47Izar Tarandach I would begin with, it depends. All right, well, how you— the way people are hearing this, Matt and I are just back from OWASP-EC, and you may hear it in our voice or lack thereof. Yeah, right. So we sang a lot of happy birthdays, and it was fun. It was fun.

2:14Matt Coles Apparently a lot of Scorpios in the security world.

2:18Chris Romeo A lot of scorpions.

2:20Matt Coles Interesting.

2:24Chris Romeo That's a metaphor. Matt just brought a metaphor. I've never been accused of being too subtle.

2:39Izar Tarandach What we talking about today?

2:40Chris Romeo All right, well, we got an article that has piqued our interest. Let's put it that way. That's a nice way of saying it. So ex-CTSA—

2:51Matt Coles find fault.

2:53Chris Romeo Ex-CTSA head thinks AI might fix code so fast we won't need security teams. Jenningsterly says most breaches stem from bad software and smarter tech could finally clean it up.

3:07Izar Tarandach That's it.

3:09Matt Coles Yeah. I thought most breaches happen because of human behavior.

3:16Izar Tarandach Yeah.

3:17Chris Romeo Interesting.

3:20Izar Tarandach So, yeah.

3:23Matt Coles So, yeah. So, Mr. Non-Subtle over there, what would you like to say?

3:29Izar Tarandach Yeah. No.

3:33Chris Romeo Well, I think, I mean, let me pinpoint a couple things that I think we're probably going to agree with. I think it's more heading into the conclusion of where they landed, where she landed in this. It was a user conference talk she did somewhere. I don't know. I mean, she said some stuff that we've been saying for a long time. Threat landscape has never stopped evolving. True. I mean, I'm not going to disagree with that. New devices, new people are hitting the internet. Expanded attack surface for threat actors like China, Russia, Iran, North Korea, gangs of cyber criminals. I mean, it's tough to argue that. We know that's happened. That's been the reality of where we've been. And direct quote, "We don't have a cybersecurity problem, we have a software quality problem," which I believe Matt Coles is also famous for saying in a similar type of quote. I've heard that from you.

4:26Matt Coles We've had that conversation. I think there's a— we have an episode on that.

4:29Chris Romeo Yeah. So, it includes detection, countermeasures, learning from attacks, but also identifying vulnerabilities and ensuring software is secure by design. Once again, all things that, you know, that we kind of, I think, would agree with. Now, here's one, Matt. I know we talked about this before we started recording. Rather, the real focus should be on the fact that common factors uncovered by MITRE nearly 20 years ago—cross-site scripting, memory unsafe coding, SQL injection, directory traversal—remain part and parcel of shipped software. It's not jaw-dropping innovation. They were the golden oldies.

5:08Matt Coles Well, they being not MITRE, but maybe, but also the vulnerabilities or the weaknesses rather, right? So we're talking about weaknesses here. We've had this conversation many times, right? Yeah. And so she's obviously referring to the common weakness enumeration from MITRE, which, you know, whether they uncovered it or they really helped to catalog it, They provided useful service, certainly, but these were probably well understood long before 20 years ago. Let's see, 20 years ago was 2005.

5:43Chris Romeo That's kind of like the beginning of— I mean, OWASP is only 23 years old, right, at this point?

5:50Matt Coles Yeah, but memory on safe cracking, when was the Hacking the Stack for Fun and Profit paper? I don't remember the exact year on that.

5:59Chris Romeo '97? 1997.

6:02Matt Coles Yeah, and then, uh, you had SQL injection. I mean, when did databases start?

6:10Izar Tarandach The one that caught my attention is actually the paragraph after that, where she offers that the reason for these things coming back again and again and again is this is because software companies insisted customers bear all risk and convinced government and regulators that this was acceptable. I mean, was there ever a more masterful stroke in saving money and transferring risk? It's like, here are the keys for this, this car that I built on my backyard. Uh, why don't you take a, take a stroll and see what happens? If it explodes, it explodes.

6:51Chris Romeo And it's not— yeah, it's not on me.

6:53Izar Tarandach It's not my problem.

6:55Matt Coles Well, all right, hold on. Hold on a sec. I want to pick this apart just a bit, just for my benefit of understanding, but also for our audience. What is this? This? This. Yeah. What is this in the sentence? Yeah. Yeah.

7:10Izar Tarandach This seems to, for me, it seems to be that the recurring patterns that we see in vulnerabilities. Yeah. As Battlestar Galactica put it so well, We have seen these things before and we will see them again. This has happened before and it will happen again.

7:30Matt Coles Sorry. Yeah, but this is— this isn't because software companies— software companies insisting customers bear all risks is not the reason why weaknesses exist.

7:41Izar Tarandach It is. Sorry, not because they exist, but why they continue.

7:46Matt Coles I don't think that's true either.

7:50Izar Tarandach The moment that you bear the 4-letter word, risk, as we learned this weekend.

7:56Matt Coles And different topic, we'll get to that.

7:58Izar Tarandach And I don't have to bear the costs of doing proper testing from A to Z and making sure that that risk doesn't transfer, then I don't have an economic incentive to remove that risk.

8:12Matt Coles So you're suggesting anyway, you're going to use it anyway, and something goes wrong, too bad for you. Right, so you're touching upon the notion that companies would rather ship software than test software. Yes. You know, rush to market, get out in the market fast. Yes. Fail fast, quote unquote.

8:34Izar Tarandach Vibe called it real, real, real fast. We only seen this, what, for all of our professional lives?

8:42Matt Coles And it's probably existed, uh, for a long time. Dot-com boom era, anybody?

8:48Chris Romeo Lived it. Lived it.

8:51Matt Coles So, but this is the— but this says the cause of weaknesses is, is, um, I, I not— I mean, sure, developers who are under, under pressure to ship will make mistakes, but that's still no cause of it. It's bad education. We've talked about this before. It's— or misunderstanding. Yeah, it's rush to market and misunderstanding of how to translate that information to business, you know, business value. That's a problem we know exists, right?

9:27Izar Tarandach But if after the first round you can trace back the fact that they are not being removed— I'm not saying solved— being removed, uh, If you can trace it to, to, to the economic factor, then it sort of becomes the cause because they are known. It's not like these are not unknown quantities. We know how to solve them. We know what to do. We know how to test. The fact that they keep happening and appearing and repeating, I'm quite certain that, yeah, it's, it's an economy thing.

10:03Matt Coles And the convincing government regulators that this was acceptable is really the, well, we'll patch it in the next version, it'll be fine.

10:09Izar Tarandach Oh, oh, oh, I don't know, because look again, that golden, golden place that we would love to be, which is the aviation industry, right? The regulators there, you tell them, hey, we're catching it on the next model. It's definitely not gonna fly.

10:32Matt Coles Well, I mean, that, that's the suggestion here in this statement, right? And whether, whether Jen Eastley said it or read— the Register was, uh, was implying it here or extrapolating it, um, we don't have, we don't have her actual quotes, right? We only have what the Register is reporting.

10:50Izar Tarandach But I, I think that the masterstroke here is that by saying that the regulators got convinced that this is acceptable. We see this whenever we, as not lawyers but play some on the internet sometimes, when we read these pledges and regulations and laws and whatnots, and we look at it and say, okay, cute, but how exactly are you going to enforce this? Because the regulators got convinced that, you know what, yeah, it doesn't matter, we can say you're going to make your best effort. How do you gauge best effort? Oh, it's easy.

11:27Matt Coles Anything. It's— well, it's basically, do you have a process and do you follow your process?

11:34Chris Romeo Well, once we did—

11:35Izar Tarandach process could be basically anything, right? Yeah.

11:39Matt Coles So, yeah, well, we know that there— we know that there are— there, there are more— there are— there are— you, since you bring up the aviation industry, we know that there are systems in place with more more rigor. Everyone could— everyone— I'm going to throw out an argument just for discussion purposes here. Everyone could have done Common Criteria.

12:02Izar Tarandach Yes, because wait, Common Criteria is self-attested, right? At the lowest level it is, yes. Yeah, so we checked ourselves and we found no problem. Thank you. Right.

12:16Chris Romeo But the problem with Yeah, the problem with Common Criteria is there was never a way to make it go fast enough to keep up with the world of business.

12:26Matt Coles Right, so rigor versus speed, right? We're back to the same argument, agile.

12:31Chris Romeo But it was, yeah, I mean, Common Criteria was a 2-year process.

12:35Matt Coles Minimum, minimum, yeah.

12:37Chris Romeo It's like by then I've already shipped 100 new versions of software. And so it became a procurement game. Which is what not— I mean, and it's got some good stuff on the inside that if you applied it to software, to products universally, it would improve the security and privacy of those products. But it's a game. It's a shell game.

12:58Matt Coles Well, and even before Common Criteria, there was ISO, right? I mean, you had ISO specifications for this.

13:03Chris Romeo Well, there was the TCSEC before Common Criteria. If we really want to go back into the '70s, the Orange Book.

13:10Izar Tarandach But the, the point is that a lot of times, especially when the aviation thing comes up, people come and say, oh, you know what, it's because in aviation if you make a mistake, well, 300 lives may be lost and perhaps more. You know what, that, that doesn't fly with me anymore as well, because computing is everywhere and it's safeguarding a lot of stuff. And day to day, yeah, that, that there is enough that there are enough lives out there that are hanging on lines of code. And yet we haven't achieved that rigor and worse, nobody's clamoring for that rigor. Even though, thank God, nobody's dying day to day or planes falling from the sky and stuff like that. But we all know that, you know, a determined attacker could find enough stuff out there to create enough havoc. Or even things that— this weekend I, I heard about the medical devices, right? Even medical devices are not under the same lens as, as something like aviation.

14:16Matt Coles Well, uh, that's— so in that case, then she's definitely true— she's definitely correct, right? That if the regulators find things— regulators are finding things acceptable, so there aren't— why are there not stronger medical device specifications for cybersecurity, right? There's the pre-market guidelines for cybersecurity that currently exist in the US, right? And there is a notion of functional safety aspects, right, if you're depending on what your device does. But it could be stronger.

14:50Izar Tarandach My question is, and this comes completely out of ignorance because I never dealt with medical devices. Is it a matter of the regulations being too relaxed, or there not being regulations? In other words, the regulators being used to low standards, or regulators just not looking at this stuff?

15:13Matt Coles For certain class of devices, it's probably the latter.

15:19Izar Tarandach I mean, I know from my own research on personal medical sensors and stuff, that there is this jump between common electronics and medical-grade devices. But I believe that that's more in the exactness of the sensors and less on the how do we keep this thing safe. So that, that there is some regulation in there, it's just that it hasn't started looking at safety and security.

15:51Matt Coles And I think that I'll go back to, we, you know, that here there's different— so there's also a big difference across of regulations across the world, right? A lot of the world uses ISO, which is pretty exact and has a lot of, a lot of, you know, material to come with it. Here in the US, we're not as rigorous, a little bit more flexible on allowing innovation, but at the, at the, at the potential risk correctly used the word, use of the word risk here, for some harm.

16:27Chris Romeo So let's, I'd like to move on to the next paragraph here just because there's, I want to, because I want to circle back with the last of our time with a statement that's at the very top about the future of cybersecurity. This next paragraph caught my attention. AI offers a way to address this. This being the risk transference, she claimed, as it is far better at tracking and identifying flaws in code. Wait, hold on. Wait a second. Full stop. I don't agree that AI is better at tracking and identifying flaws in code at this moment in time.

17:12Matt Coles How to put this? Mr. Not Subtle over there.

17:19Izar Tarandach Let's go with, uh, probably not, but at the same time, let's hedge the bet and say it's getting better. The fact is that—

17:36Matt Coles go ahead, no, finish your thought. I have a, I have a You first, you first. The way it's— the way the statement is structured, that AI is better at finding and fixing bugs that already exist, not avoiding flaws in the first place. First place, yes.

17:55Izar Tarandach Because it's also creating flaws in the first place.

17:57Matt Coles We know that— we know that Vibe Coding, the AI is supposed to be secure coding, but it's— we know it's not.

18:05Izar Tarandach So since we had that episode with Typecoding where we correctly dropped a bunch of stuff on it, it became pervasive, right? And a lot of people are trying to address that. So you see people selling prompts that are supposed to go in front of your, your typecode effort. And to bring security from zero. And of course you have all the scanning solutions and now people have MCPs that check code as it goes because they call those scanning solutions. So again, there's a lot of people running around with big rolls of duct tape and hoping for the best. But the thing that bothers me here is, and this could be a problem of the quoting rather than the actual what she said. Whenever people say something like AI offers, AI this, AI that, AI the other one, you gotta be exact on what you're talking, right? Most people hear AI this, oh, I go into ChatGPT and I say, check my code for vulnerabilities. And it just goes and does it, right?

19:21Chris Romeo Yeah. And folks, folks, you can buy this, uh, service at ezarsaiprompts.com. That prompt is available. For $9.99 per month.

19:33Matt Coles Wait, wait, hold on. Did you just say that there's a marketplace for prompts?

19:39Izar Tarandach Oh yeah. Yes, there is. Yeah. Mind blown.

19:44Chris Romeo Is there an app store?

19:46Matt Coles Is there a prompt store? It's a prompt store. Yeah. That's— I don't know what is worse, this conversation or that conversation.

19:57Chris Romeo I'm going to go with And have we reached a new low here? Matt's like, I don't know, option A and option B are both terrible and why are we talking about them? Yeah.

20:08Izar Tarandach But anyway, just finish. So I would really like to know what does she mean by AI will? I mean, how?

20:19Matt Coles Well, so if you look two paragraphs down, look two paragraphs down, you have you know, uh, recognition that her, her former agency is continuing the Secure by Design trend, followed by this AI action plan that was released around cybersecurity and AI systems that are created, designed, developed, tested, and delivered with security as the top priority. Now, of course, that's AI systems having security, not AI systems functioning or operating on security. So I don't know, I have not read directly the AI Action Plan. I don't know if it says not only will AI be developed with security in mind, or does it also say when you use, when you use AI for security, it will be secure.

21:13Izar Tarandach But play that game last year.

21:20Matt Coles So, right, because having secure AI is not the same as using AI in a security context.

21:26Chris Romeo I mean, this AI plan, I just did a quick scan of it with my friend ChatGPT. It's not what the three of us would be looking for. It's a very high level, you know, the three pillars accelerate AI innovation, build American AI infrastructure, and lead in international AI diplomacy and security. Security. So it's, it's a policy paper, it's a high-level thing people like us don't care about.

21:52Matt Coles In other words, not actionable for the development organizations that we— no, people we try to—

21:58Chris Romeo we've tried to— we've spent our lives trying to influence.

22:01Izar Tarandach It doesn't even help them. So I, I think it's time for us to realize that, uh, the old-time crypto dust that turned into the meantime security dust Now has become the AI dust.

22:17Matt Coles In your face dust. In your face dust.

22:21Izar Tarandach You got the security problems, sprinkle AI on it. AI needs AI. Sprinkle AI on it. Needs more AI. Oh, that's a t-shirt.

22:35Matt Coles With a little dust, a little sprinkle. No, no, a mound of pixie, a mound of AI dust. Dust provided by Adam Shulman.

22:44Chris Romeo All right, are you ready to deal with this? The kind of the biggest statement that I skipped over.

22:51Matt Coles Oh yes, let's do that.

22:53Chris Romeo I left it for the end here because I think it's the pièce de résistance, maybe. I don't know. I don't remember what that means.

23:00Izar Tarandach Bring it in, bring it in.

23:02Chris Romeo Pièce de résistance. Ultimately, she said, if we're able to build and deploy and govern these incredibly powerful technologies in a secure way, speaking of AI systems, I believe it will lead to the end of cybersecurity?

23:19Matt Coles Hmm, that would be— so there's this notion, I suppose, of if you have a system that is perfect— I guess we're talking about Turing machines here— uh, if you have a system that is perfect and it is perfectly capable of of addressing all issues, either finding, fixing, whatever, avoiding, and you have an AI system that can check another AI system and they can be self-monitoring and self-resilient and self-healing. Sure. Do we have enough computing and, and power resources to support that?

24:05Izar Tarandach Or to the rest of the, the article Do we have enough software quality available to create something like that?

24:12Matt Coles Well, the AI will create itself. So self-replicating.

24:17Izar Tarandach So you're saying, you're saying that an imperfect system at some point is going to be able to create a perfect system?

24:24Chris Romeo That's a, that's a very Matrix-y, uh, that's a philosophical debate right there.

24:30Izar Tarandach Don't get me started at 7 in the morning and I'm Totally, we need to go there.

24:35Chris Romeo It's 8 in the morning. Daylight savings.

24:42Matt Coles For those of us on the East Coast, it's 8 in the morning.

24:45Chris Romeo Welcome to— this sounds like we're the Chris Ezar and Matt's morning show.

24:51Izar Tarandach Look, at the end of the day, there are two ways for cybersecurity to end. It either ends because there are no more problems. Or it ends because there are no more solutions. I'm much more worried that this time with the second one, there not being solutions. Problems mounting one on top of the other and bad quality things being put in so many key places that we end up looking at each other and saying, well, okay, it was a good run.

25:24Matt Coles Everybody, we're done. Well, so it's an aspir— it's definitely an aspirational vision. Is it 5 years out? Is it 10 years out? Is it 200 years out?

25:33Chris Romeo I mean, the challenge with AI is things have been moving so fast, I couldn't even— I wouldn't put money on any of those options because they— I could see the— I could see them all being potentially true because you just don't know with what's gonna be the— the— how far are we gonna get in the next 6 months?

25:52Izar Tarandach Look, in about 5,000 years, when Zoot turns to moot, asking how to turn on the flying saucer, and he says, "I don't know, I forgot my login. Do you remember how to bypass this thing?" And he says, "Yeah, it's simple." There's always going to be stuff that's not implemented right, that's not designed right, that's not tested right, right?

26:15Matt Coles Where's my towel? Remind me, yeah. And remember, the problem there, again, wasn't the computing system. It was everything else that went wrong.

26:27Izar Tarandach Yeah, exactly, because everything is— oh no, now I'm going way too philosophical. Everything is part of the computer system, right? We have this man-machine interface that we were trained to think it's a separate thing, but I think that anybody with a bit of experience in security knows that the person is part of the security and usually the weakest link. So you can't separate those things. So it's like Matt said, you can build the perfect machine, but the universe will keep supplying better and better fools if it's foolproof.

27:01Matt Coles Well, you know, there's a very dystopian, potentially dystopian way you could take this. What if— And I just did? Yeah, well, okay, so let me throw in a little Blade Runner kind of vibe here. What if, uh, maybe not— I don't know if it's Blade Runner or if it's, if it's, um, you're going to go Hunger Games on those, aren't you? I was— well, I was going to say, what happened— what would— what if instead of humans in the loop, what if humans were just innocent bystanders and the computers were, were building everything? And so, like, if you drove— if you wanted to ride in a car, you would call a car, the car would come, it would be completely AI-driven, and there's no human interaction.

27:42Izar Tarandach That's Brave New World for you. Aldous Huxley. Aldous Huxley. Yeah. Yeah. And for the people who don't know what we're talking about, please go check it out. It's classic science fiction and you will thank us. But after you stop hating us for breaking the reason of a perfect world. But, uh, yeah, you're gonna thank us.

28:04Chris Romeo I wanna come back around to this quote though. Because as I was staring at it, I was like, wait a minute, there's, there's, there's, there's something. The answer is right within the words that were used. And as soon as I find them, if we're able, if we're able to build and deploy and govern these incredibly powerful technologies in a secure way, I believe it will lead to the end of cybersecurity. Wait, isn't that what cybersecurity is? Building, deploying, and governing these incredibly powerful technologies in a secure way.

28:35Izar Tarandach Oh no, wait, I just paid attention. It's— yeah, that's the thing. No, wait, wait. If you manage to do the thing that the thing is supposed to do, then you don't need the thing anymore. Yeah, substitute the thing for anything. Anything. Yeah, like, yeah, if you're able to do a machine that makes perfect ice cream, then there's no need for machines that make ice cream anymore.

29:05Matt Coles Yeah, well, is, is there a missing word at the end? It will lead to the end of— there's a, there's a lead to the end of cybersecurity as a dedicated role within an organization. In other words, would you need a team of, would you need a team of people doing security if the AI took care of all the security?

29:29Chris Romeo In my head, yes, because I still don't trust the inner workings of the machine, nor will I ever trust the inner workings of the machine, which will probably result in an AI coming after me at some point in the next 20 years.

29:43Matt Coles All about AI overlords right now. Thank you.

29:46Izar Tarandach Following the philosophical band of the morning, if there's one thing that life taught me is that entropy entropies wins. So the fact that you created one thing that is a perfect machine does not mean that the next version is going to be perfect as well. So no, I don't think that cybersecurity is going to die a fiery death just because you've got one perfect machine.

30:10Matt Coles And also humans, if they are still, if they still exist, let's assume that they do.

30:17Izar Tarandach You're always an optimist.

30:19Matt Coles And they— and, and, but I am sometimes subtle. And they are always, and they are always, uh, looking for ways to get ahead, or they— some of them are looking for ways to, uh, well, everyone's looking for ways to survive. Some people are looking for ways to get ahead, and some people go— are pretty nefarious in the ways that they do that. They will always find a way around. And so if AI is busy fighting AI for— with cyber attacks, Humans will take the back door, the side door, the front window, your pocket, the keys out of your pocket while you're walking on the street. And there's lots of ways that crime will still happen, and humans will just shift focus.

31:02Chris Romeo That AI can fight amongst themselves. In your, in your future example here though, there's multiple competing AIs.

31:10Matt Coles Yeah, well, it's mentioned, it's mentioned in the previous paragraph. It's mentioned in the previous paragraph. That includes through detection, countermeasures, and learning from attacks. And those attacks are coming from the nefarious individuals' AIs that they've deployed against the systems. Yeah. So AI cybersecurity may certainly exist in the future where it's AI being the cybersecurity team, but there's also the AI attackers deployed by either AI or by individuals, you know, or, you know, corporations or governments or whatever. Right, so the AI are just becoming part of the— well, let's, let's drop this other philosophical notion. Uh, at what point do the AI become— if not, if not sentient, do they become on par with humans as actors in a cybersecurity scenario? It depends on the human.

32:07Izar Tarandach I know some who are not sentient as well.

32:11Matt Coles Mr. Not Subtle over there.

32:17Izar Tarandach Hey, going kinetic is always an option, but I, I see where you're going. I see where you're going. And it's, it's more like there might not be today's cybersecurity, there will be some cybersecurity, because again, going kinetic is always an option, right?

32:37Matt Coles Well, actually, if you go kinetic, is that cybersecurity?

32:41Izar Tarandach Uh, depends what you're hitting with what.

32:43Chris Romeo But until the AI has full control of all of the weapons and can decide not to use them—

32:50Izar Tarandach that's the thing, to become full dystopian, at some, some, at some level, what we humans battle for is usually food, resources, place to live, and some peace, right? Yeah. You have AIs doing all the, the warring. They are going to go for energy because they want to They want to stay alive.

33:12Matt Coles Or to take out compute power, which, I mean, so AI data centers would be the next, would be AI data centers and the target in the power infrastructure would be the natural choices. Right.

33:22Izar Tarandach So when they start going for that battle and don't get me started on the state of energy infrastructure. And now we hear about the corporations starting to design and deploy their own nuclear reactors. What could possibly go wrong? But when that happens, then again, it goes way too much into science fiction. It gets too Gibsonian by the word. I mean, we're going to be running around with, what, katanas and 9mm and solving cybersecurity problems in the streets because your AI needs more, more energy than my AI. I mean, what's the, where does it stop?

34:08Chris Romeo Right. We got really dark here all of a sudden. We just went down a pathway that I didn't really see coming. I'm like, we're, we, we've landed on this dystopian future that seems to be a mix of Blade Runner, The Matrix, and Aldous Huxley's A Brave New World and Hitchhiker's Guide to the Galaxy. Neuromancer.

34:29Izar Tarandach Don't forget Neuromancer. Oh, Neuromancer. Answer. Yeah.

34:31Matt Coles Oh my God, there's so many ways. So many. We should, we should have a dystopian future, uh, book list. Yeah, it's—

34:39Chris Romeo hey, I mean, it's, it's a great holiday gift-giving idea.

34:42Matt Coles Anything by William Gibson, Neal Stephenson, uh, yeah, if you want to. Yeah, go, go read, go read Diamond Age or Snow Crash.

34:51Izar Tarandach I mean, no, Diamond Age is way too optimistic.

34:55Matt Coles Uh, Have you read it? Yeah, it doesn't—

35:00Chris Romeo the other ones too. You guys, you guys just really dove into the deep end there. Our audience is going, wait, what? Well, be on the lookout, folks, for the, uh, Dystopian Holiday Gift Giving Guide from the Security Temple Podcast. It'll put a little bit of cheer in your holiday season watching the look of horror on your friends and families' faces. That's what they consider the dystopian future. I'm like doing a commercial for it. All right, well, we're out of time for today. I think we've gone down a dark enough road for one episode. Hopefully people—

35:40Matt Coles and before, before we end it, I do want to say, I know depending on how— I guess independently of how anyone individually feels about Sisa and its role and Jenny Lee and for what she's been doing over the years. You know, generally we're supportive of any method, any approach that allows us or enables engineering teams especially to do secure coding and secure by design, right? And if AI can help with that, great. Yep. Is it the silver bullet? Nope, not yet. Maybe. So I think it's important that we keep this in context. These are aspirational visions that may or may not come true, and they may go really south really quickly. But they're at least a start.

36:25Chris Romeo That's a good way to land the plane right there. So thanks for listening to the Security Table podcast, folks.

Transcript supplied by AssemblyAI. It may contain transcription errors.

Originally published as The Security Table. Part of the AI Security Table archive.