EP269 Reflections on RSA 2026 - Beyond AI AI AI AI AI AI AI

Security. Hi there. Welcome to the Cloud Security Podcast by Google. Thanks for joining us today. Your hosts here are myself, Tim Peacock, the Senior PM for, well, a bunch of stuff in Google SecOps, and of course, Anton Chuvakin, Senior Staff in Google Cloud's Office of the CISO. You can find and subscribe to this podcast wherever you get your podcasts, as well as at our website. cloud.withgoogle.com/cloudsecurity/podcast. If you enjoy our content and want it delivered to you piping hot every Monday, please hit that subscribe button on your podcasting app of choice. Anton, we are doing our annual RSA recap. This is RSA 2026. Hard to believe we've been doing these since 2021. Is that right?

No, probably— did we do? I mean, we should probably put all the RSA episodes in the show notes for this show.

Yes, we should.

But I definitely remember '23 fondly, and somehow I don't remember the '21. So either we didn't do it or something else happened, or my memory It could be your memory. I can be in denial.

That's right. I would love to be in denial about how many RSAs I've been to, listeners. It is exhausting. But what did we see this year? Do we want to start with the AI framework we've developed?

Correct. And so this spontaneously was born in my, maybe in my analyst brain and the team's PM brain. Actually, both brains are kind of essential for this one. It started from me thinking about AI washing.

We saw a lot of AI washing.

AI washing is basically when people don't do AI, but they say they do it. It's—

It's like greenwashing, but for AI.

And like cloudwashing. Well, I guess, but for AI. Yeah, that's right. That's what you said. But it's also one dot on a spectrum. And the spectrum may include things like adding things—

Oh, I was going to lay out the framework. I thought you were going to start here and we were going to go through.

Yeah, I apologize.

No, no, go on. What's the least AI washed?

The least AI washed, actually, I don't think it's— I think I'm reconsidering it maybe more than one dimension because there are people who don't use AI and don't say they use it, but they don't need it. For example, I saw a vendor making encrypted USB drives. There was no AI in their messaging. I have no idea if they use AI for anything beyond productivity, but if they don't, it's probably okay.

Yeah, I don't think you need AI for that.

And so this to me is a vendor that doesn't say anything about AI, don't do it, and that's okay.

Okay.

But I suspect that if you are an AI SOC vendor that doesn't use AI, you are in the fraud business, not in security business.

And we've seen those kinds of frauds happen where people say they're AI, And they're not actually AI. And so, yes, that is a valid point on the framework to stay out of.

Yes. And also, I felt like there are people who used their 2021 marketing materials, and in a few spots, they spray-painted, and AI. So for example, we secure identity for employees and contractors. And now there's a fat comma, and they say, and agents. And they may do that. I'm not being negative on this necessarily because I didn't ask if they actually do it and how much of that stuff they really do. But it feels like to me the term was kind of like AI touch-up, not AI washing. Tim didn't like it, but it's the degree of AI use. And I suspect there's always people who use AI but don't say it.

Yeah, well, no, why would they exist? Because they have a tasteful marketing team?

Your hypothesis was that they have a tasteful marketing team.

Having once been a tasteful marketer, I'd like to think there are others out there.

But also you can do better, and Wiz did better. Wiz has— actually, it's funny, they had the main booth that was called AI agent lab, and it was very AI. But if you walk outside, you would see AI-free zone. And by the way, I'm not joking. There was this big sign that said AI-free zone by Wiz as well. So if you think there's too much AI, you go to AI-free zone.

And you drop in free donuts.

Oh, no. I missed one of those.

Human-made donuts. No AI in the donut. If you find an AI donut, tell your parents.

If you find a human donut, just eat it, I guess.

Just eat it.

Yes. So this is tasteful. Oh, how about that? I made it fun. Yeah. Look at that.

You made it fun, listeners.

But it's also on the same spectrum somewhere. People who— I also feel like there are people who pronounce themselves to be AI natives. So, the orthogonal part of this framework is that you also can be AI tourists. You just landed in the land of AI and you started using AI for something beyond productivity use cases, maybe text summarization, right? That's a baby's first GenAI use case. And in this case, if you say use AI, you'd be honest, but if you say you're AI native, you probably would be lying.

Yes.

So, the reason all this rambling that— and sort of a semi-structured framework is that when you judge a vendor by the messaging, the marketing, the glossaries, the description of their product, AI use may mean something or it may mean absolutely nothing. And you need to dig deep and you need to know what they're buying. It's very like 1990s kind of advice, maybe very classic advice, but unfortunately, it didn't change.

It boils down to get good at buying?

That's unfortunate. Get good at buying, but get good at technical information about what you're buying, right? Because some of the stuff about, say, a vendor that does AI introspection, there was a— in Innovation Sandbox, there was a vendor who claimed to be able to read the mind of a model. And know what it's thinking, only for open weights models, by the way, before you ask. And that to me struck me as absolutely fascinating from the technology point of view. May or may not be useful in security. They think yes, I don't know. But to understand how they do it, you have to have pretty deep AI knowledge. Yes.

I would have no way of evaluating their claims personally.

Exactly. Same, same. So, it is interesting that—

You know, I once put together a whole fake pitch about malware Psychology. When I was at Shape, somebody reached out for a pitch, and we put together a whole fake story about malware psychology. And now there's literally a vendor claiming to read AI minds. That's amazing. 10 years later, folks.

Yes. And this also— I mean, we can lament the hype. We can whine about it. But ultimately, if somebody is here to understand what vendor does, the genuine use of AI in their product and what they say about the AI, are almost completely disconnected. This to me is an interesting insight. And I'm going to cry a little bit, but even vendors who are in AI SOC land may genuinely not use AI for anything systematic. This to me is kind of crazy.

Or at least not to the extent they claim.

Or not to the extent a reasonable, non-AI-savvy buyer would understand.

Yes.

So, and this to me is— I think we explored enough, but I feel like it's also hugely interesting.

I want to talk about, which is The risk that was pointed out to me by a dear friend on the floor that many of today's vendors are likely to be completely run over by the AI labs.

So this is a question that gave me pretty strong flashbacks to some time when some big, big CEO honcho of some company said RSA is over because in the next couple of years, 3, 4 security vendors, large security vendors, including his company presumably, would acquire all the startups and the consolidation would happen and we'd be done.

And yet it hasn't.

Nothing even remotely similar happened. And so consolidation in security is not a thing. Well, it's consolidating, but never consolidated. I think that's another term. So the first thought I had when you said somebody believes that AI vendors would own all of security is this reaction, like, no, because no. No, because— That's not how the industry works.

But at the same time, surely you see the capabilities of the models coming along. Surely, you see how if a company adopted these tools natively, you maybe don't need a vulnerability scanner if you can run Claude on every endpoint.

Tim prodded me for a counterexample. And I said, oh, yeah, sure. Here's a counterexample.

Windows 2000.

And while I was making the counterexample, I thought, that's not a good counterexample. So, I tried again, and I kind of failed again. And I tried for the third time, and the third time was actually even worse. So, I philosophically disagree with it, but I Don't have good arguments. The only direction for good arguments would be this. I feel like there's a lot of enterprise knowledge that bad security vendors also don't have, but good security vendors do have that.

Yes. And what's funny is that's exactly what the other guy standing next to him pointed out was this actually raises the value for a vendor of knowing what bars to go to at RSA to find the CISOs that are hanging out and pitch to them in person. And so perhaps industry veterans listening to the show, your knowledge of where to get a drink at 5 o'clock on a Wednesday at RSA Week has become more valuable, even though Gemini, Anthropic, OpenAI have a real risk of coming for our market.

But the other side argument, which is kind of a spin-out from this argument connected to people debating the value of the analyst firms, like my former employer, Gartner, when people say, look, stock of such and such analyst firm has decreased because people assume that they can just go to, quote unquote, ChatGPT and ask everything they would ask Gartner.

Yes.

And the odd part is there is definitely a category of questions that people ask analysts that can be answered by an LLM with a sufficient degree of utility. So, it's there, but the problem is most deeper questions rely on analysts, large analyst firm knowledge base, fact base, as I said, which isn't public, and you can't train your AI on that because the company has that, but a generally smart machine won't have that. So I don't know how to counter that.

Is analyst fact base just a nice way of saying all of the gossip the analysts have collected during their time talking to big companies?

Correct. But integrated and aggregated. Integrated and aggregated.

You had a gossip management system, a GMS.

Essentially, yes. But it is called inquiry data. And that means basically what you've heard during the inquiry from a client, questions, answers, context. And this isn't—

That's a nice way of putting it.

All of this isn't public. And I feel like if you have a robot answering without knowing this, no matter how smart, it would often enough answer wrong.

Yes. It would not have the rument, as some people like to call it.

Correct. And so, in that sense, to me, analysts may be kind of diminished, but they would survive. So, that brings back to the—

That brings back to the security industry. It might be diminished, but we'll survive.

Yes. The diminished part is there because deep in my heart, the static analysis Yeah, sorry, guys. I'm not an AppSec person, and you've seen me say it.

We know he's not an AppSec person.

I prefaced some blogs, I mentioned AppSec by saying this, but static analysis just seems perfect for LLMs. And I don't see how quote-unquote secret enterprise knowledge would help you tell good code from bad code.

Or tell it better.

I've seen some SaaS vendors, static analysis security vendors, make passionate arguments on, well, Twitter, where else, that they do matter despite, what do we call it, cloud code security, whatever it was, the thing, right?

Or from Google, you know, Deep Sleep. Yep.

And I don't believe them. They're not credible to me. But that's SaaS, is just a uniquely clear example where AI can make a huge dent or possibly decimate it.

Surely there are other examples, like a firewall rule analysis tool. Why do I need a specialized tool to analyze firewall rules when I could use Gemini code?

Policy writing, a lot of IAM decisions. But— Yeah.

Why do I need a PAM tool anymore? Put the justification into the chatbot. Let it decide.

And people do do that, by the way. And there's also a degree of automation around that. I think that AI in a SOC has areas that are very much that way. And there are also areas which, to me, are not that way. So, for example, building detection rules in a way that matches your enterprise reality implies a deep degree of knowledge of the enterprise reality. So you don't FP on every custom-written app that you build.

But surely, the agent will figure that out during the triage phase, feed that back upstream, and voilà, a week later, you don't have those false positives anymore.

Unless it uses tribal knowledge that's in human heads, not in machines. And I think that if what you do in security relies on Copious chunk of tribal knowledge, a super intelligent chatbot would not get you soon. Later on, who knows? Maybe they build like an interview bot, right? Which can go talk to humans, collect the data, structure it.

Well, that's trivial to send a Slack or a Google Chat to somebody and say, hey, did you really mean to run PowerShell just now?

Yeah. And also ask you like why and a few other things and then analyze the data in the context. So, I don't know.

That sounds pretty easy to me.

Yes, but it may be all the information you need or not.

So, AI maybe does fit in the industry, maybe not. I want to switch it slightly and talk about AI hacking.

And yes, AI hacking, the pentesting, red teaming, and no, I'm not using the terms synonymously. I'm using them in aggregate. AI hacking, AI pentesting, AI red teaming. And this is, of course, using AI for all of those, not testing AI. on an upswing. And I've met people who are like huge skeptics. They say that if there's so many vendors, why can't I just code the same framework? That's one counterargument against the too many vendors. There's always an argument about pen testing and red teaming, that the top-shelf firms would crawl through your air ducts to get to your secrets. And I don't see robots crawling through my air ducts. Lawyer cats, maybe. Robots—

But so the scary thing to me is not just the red teaming and the pen testing, and we use those terms in the aggregate not to mean the same thing. But attacks powered by AI, like actual adversaries.

Like a bad guy with AI.

A bad guy with AI, or just a bad AI.

Oh, a bad guy with AI or a bad AI. These are actually 2 actors. Excellent point.

Those are 2 different actors. And we're seeing those supply chain compromises this week, literally, with companies getting their secrets dumped, their code dumped. I heard that a former US security leader is very pessimistic about our prospects over the next 6 months in the wave of AI attacks. I'm very pessimistic about that. And so do you have anything to cheer me up about our future, or are you also in the Heather Atkins, Bruce Schneier, Gadi Evron, AI is eating our lunches and we haven't even realized it?

This is an interesting one, and I think that you just referred people to at least 2 of the past episodes we've done on this topic. And maybe I'm weird, but I'm not losing sleep over bad guy with AI, and for sure not over bad AI.

Same word about why.

The first argument starts with, if your security is bad before AI—

It's bad afterwards too.

Yeah. Christoph had a line back in the day, like, if your on-prem security is bad, the cloud security would not change it. And I feel like if enough people have suffered from more traditional non-AI-powered ransomware, so for them, they lose today. And the fact that they lose faster with AI is not a huge change. Sure, it would lead to scaling. More of those people would lose. Somebody had luck-based security in 2005, they were okay because, well, luck. If attackers scale attacks with AI, luck-based security would be dead, would be gone. You can't have luck-based security. Yes.

Luck-based security goes away very quickly.

This is an interesting observation.

Yes.

And another reason—

Not just AI that's faster. What was the stat in the MTrends report that it's now a 30-second handoff between initial access brokers? And ransomware gangs?

Yeah, because the machines— yeah, this is like the modern supply, modern digital business in the attacker side. It's like they modernize the business using, well, computers and stuff, right?

Who would have seen that one coming, listeners? Modernizing a business with computers and stuff. Crazy. But again, that's back to the pessimism, and we're all going to be losing for a while as an industry.

I mean, to me, a lot of fundamentals produce an answer because ultimately, if you don't have cloud misconfigurations, No amount of AI would exploit your cloud misconfiguration. And if your defenses against misconfigured systems, instances, containers are robust, and they're becoming better, whether AI is helping or magic or unicorns or whatever else, it doesn't matter that the bad guy has AI. Your stuff is done well. Heather brought up an example about the applications being coded more securely. She meant with the help of AI.

Yes.

But even if you do it by magic, If your app has no obvious exploitable holes, then it doesn't matter that the bad guy has AI. There's nothing to find.

If you're secure, you're secure.

Yeah.

If you—

there's nothing to find. I mean, so to me, if we do things well, we would end up in a good spot. That's why I'm not losing sleep.

Getting the fundamentals right is a very Google answer to this problem.

It's a very Google answer. Yes.

Like a shockingly Google answer because all of our How Google Does It episodes reveal There's no magic.

In DNR, there is.

Okay. There's mostly no magic, but getting the fundamentals right produces magical outcomes.

Yes. I mean, the VM episode, 109, is basically an episode when I was at the end, I was saying, hey, but wait, where is the magic? Because I was conditioned by the DNR discussions when people whip out some magical piece of tech and everybody goes, whoa, no way. And in the VM episode, the guest basically said, that's what we do. And I'm like, That's what everybody should do. And I think she said, but guess what? We actually do it. And everybody else, guess what? Actually doesn't. On that somewhat humorous and somewhat gloomy note, I want to switch to securing AI.

Okay.

Not only from bad guy with AI, not only from bad AI, but also from bad guy without AI.

Okay.

And I feel like last year, and in the similar recap, I was lamenting that AI for security was overpowering in terms of vendors, presentations, securing AI. I was almost sad there wasn't enough security for AI. I did some discussion about shadow AI and stuff, but I was surprised. This year, I feel like securing agents, securing agent identity, identity for agents, all this stuff is there. I think people do data security for AI world, and this— I'm happy to see that. And I was right that it would come, and it did come.

And did it come from new vendors, or did it come from old vendors changing their messaging, or old vendors changing their products? And those last 2 are different options.

My suspicion is that you would not like my answer, but the answer is all of them.

Ah, that's a very analyst answer.

Yeah, because I've seen startups that do things which in my mind aren't very startup-y, like say, new identity system to support agents doing work at enterprise, which I feel like it's very difficult for somebody who doesn't own anything called Active Directory to make an impact. But data security may be a little more clear. Like, there was some, what I would call an AI-powered DLP that is, to me, quite credible. Like, yeah. I mean, not literally, because I don't buy things, but I'd buy the argument. Yeah. And I think that securing AI, securing the supply chain for AI, securing training data, kind of sort of know what to do. And some of the tech for this is coming from both startups, and startups can deliver, and also large companies. I mean, we probably have some stuff in our back pockets, deep pockets, whatever, that does some amazing stuff for us internally, and it can become a product. Who knows? And so, to me, big companies here may be a credible player. I would put it this way.

Yeah.

Like, you don't assume that it's big, bad, slow, small, tiny, fast. I don't think that works here. I feel like the big and fast— Google's pretty fast these days— can be actually quite successful. I mean, we have been like, we are putting our Google hats Very deeply in our heads, but it's also a fact. Yes.

And listeners, we try not to wear our Google hats too, too obviously.

But I mean, the thing can move.

Yeah, yeah, yeah. It can. It turns out. It turns out when it has to. I want to talk about what we were calling the whiz effect that we saw on the show floor. We saw a lot of booths, listeners, that unlike previous years were not covered in messaging buzzwords. For a long time, you walk the RSA floor and you are assaulted by things that say, Zero trust, XDR. And I think I only saw one booth of zero trust, one booth of XDR. But this year there were a number of booths that were information-free and just whimsical, which really mirrors the Whiz booth strategy. Unfortunately, the most whimsical booth I saw, I have no idea what the company does.

Even now, even now, we like 2 days later, we still don't know what they do.

We haven't figured it out.

Yeah. So whimsical booths are fun and they may draw a crowd and unless people already know you because you're called Wiz.

Yes.

And you are worth $32 billion.

Yes.

I kind of have to mention this. Your whimsical booth strategy may not succeed. Your booth would be remembered, but not your message.

Right.

I mean, last year it was goats. Somebody had goats last year. Oh, somebody had goats?

I didn't see any goats last year.

I think the goats, yeah, but I've seen the Vikings.

There was some backlash to the goats. Did you see the backlash last year?

I've seen the backlash. And so no goats this year. But this year there was like UFC fighters or whoever they were.

Yeah, I saw the UFC fighters. I saw superheroes. No Vikings.

Do you remember all the Vikings that used to walk around? Vikings, right. Yes. But the fighters, I don't know what company it was and why it matters.

But there were other Wiz effects on booths as well. There was a lot of softness to booths that I hadn't seen in the past, a lot of purples and pinks in booths that I hadn't seen in the past. And so just personally, I appreciate that our industry has moved away, and I say this ironically in an all-black room, from all-black messaging.

Yes, correct. And then, of course, from other distortions of the past that are kind of fortunately gone.

Yes.

As far as keywords and buzzwords, one of the things we wanted to mention is that some of the vendors, there were 2 types of vendors that surprised me. And this is very broad and it's very RSA reCAPTCHA and probably isn't for a normal podcast. There are vendors whose absolute peak was when Tim was in kindergarten. I'm thinking of one particular firewall vendor.

Yes.

It will remain nameless in this particular case, but it was like really cool when Marcus Ranum was younger. He just invented a firewall. It was amazing. I was in school at the time, probably. And I mean, they have a large booth. And when you ask on the street, do people use them? Presumably, your friends and my friends will say no, but clearly, somebody does, which led me down to the path of, if you are, say, antivirus or a firewall vendor number 6—

You still have a market.

You think it's insane, but you're on the list. You're number 6. Somebody buys you. Somebody pays you hard-earned money. To buy a firewall from the 6th best firewall vendor.

Yes.

Or antivirus vendor or data governance vendor.

But we've observed this for years that the worst vendors are still vendors.

Correct. But then there is no— the promised AI disruption. We are speaking from Bay Area where everybody thinks AGI by the end of the week, or if you don't adopt AI competitive with AI without compute, you die. But vendors whose heyday was in the mid-2000s, if not Early 2000s are alive and well, and I'm talking about literally antivirus and firewall.

Yeah, yeah, yeah. And what I think is so interesting, and this is my negativity is about to come out.

You have it.

I do, I do. I've spent enough time with Anton listeners that I've developed some negativity. Is if you're walking the show floor as a low-information buyer, you can't tell the difference between number 6 and number 1, except maybe by size of booth. But even number 6 had a pretty big booth. And so I have a lot of negativity about the information value of spending time at RSA. I think for the informed buyer, you already know the answer. And for the uninformed buyer, you can't possibly arrive at the answer from walking around the show floor. And so I really have gotten to the point where I question the value at all of the show floor, except, of course, to give you and me something to talk about every year.

Yes, I feel like a lot of information Density is there. And I have a fear that it's lower than some past years because I did that—

Because of AI watching?

I actually don't know, but there was a— I did try this experiment a couple of RSA days ago when I would look at the booth like super quickly and do I know what they do? Like if I know the vendor name, I mean, if I look at the vendor I know, even, I mean, I'd be like, oh yeah, obviously they do X because Well, they are well-known vendor for X. Right. But then I look at the booth and say, the booth says Y. There's no mention of X. Or the booth says Y, or the booth doesn't say anything.

Booth says negative X.

Or it just gives general, like, we stop hackers from hacking. Who was it? That lampooning people on LinkedIn about that?

Augusto Barros.

What do you actually do when you say we stop hackers from hacking, or we stop zero days, or we stop whatever?

Oh, we stop zero days. So good.

Yeah. And I feel like this is below the low information density. This kind of deceptiveness, presumably.

Negative information density.

It's a negative information density. So I don't think AI is to blame. I think it's humans.

Well, that's a cheery take. We had a discussion about the signaling effect of an open ecosystem versus a closed ecosystem.

Ooh, that is a really fun one. And it's an odd thought that I had. So, and then Tim immediately like, This is more odd than my normal thoughts. So, I'm walking past, we're walking past the booth of some vendor, and they say, we have an open data lake, open detection, open something. And I thought, and I had this weird cynical reaction, is that the only reason they say open, because they kind of suck, and they're number 3 compared to a closed guy. And instead of saying, we are better, They say, we are open. And so, it is not— I'm not against openness, open source. And I'm just, in this case, I sense that promoting openness was a signal that they can't promote goodness.

Ah.

And instead of saying, no, we'll beat them in a fight, we're better, they say, they're proprietary, they're bad, we are open. But if somebody says, whose works better?

They would lose.

Yeah, it's the other guy. That's my fear. And I mean, please save me because I don't want to have this negative stigma about openness.

Well, I think openness has inherent value, positive value, strongly positive value from openness. And I think there's also possibly a signaling effect of openness is I'm so confident in my value, I will choose to be open. And there is perhaps a defensiveness that comes from being closed of Don't look too close, or I won't play with them because I want you to stay inside my garden. Because if you leave my garden, you might see how nice it is outside.

I really don't want to bring up the open IDS rules, open SIEM rules.

Oh, goodness gracious.

Sorry.

Sorry. I'm getting a phone call from the 1990s. They want their argument back.

No, I got beat up by your bosses for that argument at one point. So, yeah. So I'm not sure. Gently.

Gently. We only beat people very gently.

Oh, this is Google, right?

This is Google. You have to beat them with a smile on your face.

Yes. And I think the openness in terms of detection content is a clear positive. Like, I don't think the proprietary detection rules implies they're better.

No, I think it implies strongly that you're trying to hide something.

That's pushing it. I think it's more nuanced because ultimately nobody demands to see the antivirus rules from an antivirus vendor.

Strange.

Strange. But people do demand to see them.

They maybe should.

Maybe they should. Then they would say it's all the mail, so you can't see it. So it's a little— this is like a 2010s argument.

Yes, you can look inside the box. The contents are black.

Yeah, there's maybe a Schrödinger cat in there.

Well, let's not kill any cats by opening boxes. Do we have other RSA Recap things, or are we nearing the end of the show?

I think we are kind of towards the end, but I think that the repeated theme from all past RSAs is that the old is alive.

The old is alive.

Is very— The dream of the '90s is alive. The dream of the '90s is alive. And I feel like there's people who come and basically look how their favorite vendor of Well, firewalls doing, and they confirm that they're doing okay. And the whole AI thing just passes them by. I have no idea how it works, but I had a chat with somebody before the recording of the show, and they said there are actual buyers who are kind of largely ignorant of this whole disruption due to AI. And I said, no way. And he said, yeah, look around. There are people who are in panic. There are people who are kind of in this more or less coping mode, but there are people who are actually unaware that all of it is going on.

Yes.

Basically say this AI is a toy. This is utterly insane. It is utterly insane, but it does happen.

So, I mean, I got you in a Waymo. It's not a toy. And I think your first Waymo was either last year or the year before coming here from—

Yeah, that's right.

Yes. So, so 2 thoughts on disrupt. One, I think I will bet today that in 5 years RSA looks markedly different.

5 years?

In 5 years, okay, the IAM space and the SIEM space will be completely different.

I would do that because to me, 5 is long-term. By today's standards, 5 is long-term. And if you say 1 and 2, we can debate. 5, I would just agree.

Okay, 5, you just agree. That's too bad.

This is RSA 2031?

Yes, RSA 2031. Listeners, let's see if we're still here in RSA 2031. But 2, I think there's one more thing on disruption, which is a growing acknowledgement that it's necessary for companies to disrupt This was a topic that Sandra brought up.

Yes, Sandra.

So can we talk about that just briefly?

Yes. It's the— when I saw the announcement, I thought, yeah, people have been doing this for a bit, right? For a while. I mean, notable security vendors do try to take the fight to the attacker. But I think that nobody with our powers have tried it before.

And we've started to do disruptions. And there have been some big takedowns that Google has organized. The other space where I personally would love to see this is against the pig butchering scams. I had a long conversation with someone about that this week, and the scale of these attacks is growing. The regulatory pressure, particularly in some of the ASEAN countries, is growing. And so I think we will see more active measures there as well.

What's the polite term for pig butchering scams? There's some kind of— pig butchering scam is the polite term. Okay.

Yeah, I don't recall.

There was some kind of a discussion.

Is there a nicer way to say that, listeners? You should let us know in the comments.

Whale butchering? No, I think it's different. Ignore me. Yeah, I think that's different.

We will ignore Anton on that one. So I guess traditional closing questions then. Do you have recommended reading? And do you have a piece of advice to help people next year actually notice that the AI wave is happening?

Okay, so I think we mentioned a whole bunch of episodes and obviously past recaps. I think these are useful listening.

The Rutten-Sibbel episode, the Heather episode, the Bruce episode.

Hearing RSA recap from past years may be surprisingly cathartic.

Oh, that's interesting. I'm going to go listen to those myself.

I'm going to rely on my WebCoded app to do it. So, sorry. No, I feel like I need RSA pondering mode to be coded. I think that's a really good idea.

Okay. That's a good idea. Recommended reading?

Recommended— the tip. The tip. The reading, I think, is the—

Is the episodes.

Yes. The tip is this. If you show up at RSA and you talk to vendors, demand that they bring data. It is uncannily similar to my advice from last year, but it was contextualized to AI. So, this year, I would apply it to everything with AI. If you say AI makes it better, you gotta show the numbers. If you don't show the numbers, people must assume you are not credible and you're essentially lying. I'm sorry, vendors. We're friends. We are sort of a vendor, but to bring the data. If you say AI helps IAM, Give me the numbers. AI helps firewalls? Give me the numbers. AI powers your AV? Give me the numbers. AI SOC? Give me the numbers. I'm sorry, but like the— yeah.

I did see a firewall vendor who I won't name say AI firewalls at the top of their booth. Yeah. And I really wanted to know what that was about.

Sure.

I had better things to do than find out.

This is the perfect vendor to ask, okay, so give me the numbers. Show how AI helps your firewalls. They may have an answer. I mean, I don't want to lampoon them because it's possible there's Firewall rule analysis for a rule stack of 50,000 rules, perfect tool.

Which nobody ever wants to touch.

Yeah, but the AI may figure it out. So, I'm kind of sort of positive on this.

Yeah, could be.

But you got to bring numbers. Bring the numbers, bring the data, bring proof. Do not have, oh, we have this customer who kind of is happy because we bribed them, and that's why we have ROI. Like, get out of town. Sorry. You need to bring data, solid metrics, how AI improved Whatever you do, I agree with that advice. Perfect. That works well.

Okay. My tip and my reading is to check out the M Trends report, the CrowdStrike report, and did the Verizon report come out this week?

Week? No, I don't think so.

Last year it came out this week. So whenever it does come out, check that out as well.

And Cloud Threat Horizons as well. Ours, right? Yeah.

Not just the M Trends, but also the Cloud Threat Horizons. And with that, listeners, thank you very much for joining us for this year's RSA recap. And now we are out of time. Thank you very much for listening and, of course, for subscribing. Please subscribe so you can get new episodes piping hot. And if you love our content, please drop us a review on your platform of choice.

You can find this podcast on YouTube, Apple Podcasts, Spotify, or wherever you get your podcasts. Also, you can find us on our website, cloud.withgoogle.com/cloudsecurity/podcasts. You can argue with us on the Google Cloud Security Community site, googlecloudcommunity.com.

You can follow us on X, x.com/cloudsecpodcast, tweet at us, email us, argue with us, and if we like or hate what we hear, we can invite you to the next episode.

See you on the next episode of the Cloud Security Podcast by Google.