OpenAI’s $20B Ambition, Europe’s AI Boom, and Code Security Fears | TSG Ep. 964

Hey everyone. When did $20 billion become a drop in the bucket? In today's AI world, it is.

You're watching Textron Gang. Hi everyone, it's Alan Shimmel here for Textron and Textron Gang. You know, we, I'm up in Atlanta, as is Guy Corio who's joining us today on the gang, and a lot of our crew is actually en route to Atlanta.

So we, we have a thin gang today. But, you know, it's, it's a question of quality over quantity. And, uh, let me introduce you to our gang members.

I, I already said Guy is with us and joining us, not from Atlanta. Unfortunately, though we'd love to see her in person here. Kate Scarsella.

Hey, Kate, how are you? Doing well, doing very well. Thank you.

Actually, well, you're, you're up north. I was gonna say, you don't want to be in Atlanta. It's like 32 degrees here.

Um, but it may be just as cold where you are. It's not 32, so, Oh, okay. Yeah, then you don't wanna be There.

Well just give it a minute. Give it a minute. Well, yeah, I, I hear it's gonna be 40 degrees down home in Florida today, talking to my wife earlier, so it's cold, cold winter's here.

Winter has, winter has come. Uh, um, so guys, we've got some interesting stuff to talk about. Surprisingly, none of it Q con related, but we'll be doing cube con all week.

You'll, you'll have, uh, uh, cube con related text and gangs tomorrow, Wednesday, Thursday, and Friday. So let's stay off the cube con for now, though. I'll, I'll give you a quick thing, but Guy, our first story today involves, uh, our friend Sam Altman.

And I really gotta give him your friend, Not my friend. You know, he wins the Steve Jobs award, though. All he needs is a black turtleneck because he's well, Minus the product talent, but yeah, continue.

Oh, absolutely. But in terms of pulling the strings and playing the markets like a Stradivarius, he, he's pretty, he's got pretty damn good at it. Um, anyway, he's out here touting that AI, open AI is gonna do $20 billion in, in revenue.

And you know, what, three, four years ago, we would've given a crap. But in today's world where we talk trillions, 20 billion ain't cutting. It Feels like 20 million, doesn't it?

Yeah. Yeah. It really does.

It really does. 4 trillion. Yeah.

What do you think? Well, let's take a, just a quick step back. I'm not known for being quick, but I'll give it a shot.

5. But how do they do it? They did it two ways.

One was the really important transform, um, which was the trick that allowed for the training that you need for this, uh, AI that, uh, looks and smell, not smells, but looks and feels like chatting with a real person. Um, so that those, so the transformed trick was really important, but then they threw gobs and gobs and gobs and gobs of data added the whole internet proverbially speaking, they threw the whole internet added to train, but they couldn't do that without, I'm gonna say the word, the N word, Nvidia, Nvidia, Nvidia, GPUs, and many of them. So, fast forward to today and using this method, ology of throwing gobs and gobs and gobs of data at a whole ton of compute.

That is the model being followed for training foundational models. Sorry, I use model in two different, that's the paradigm. Are there others?

Probably deep seek probably showed another way. There are definitely other ways to train. I'm a big fan of data curation, but none of this seems to move markets in any way.

And so Altman's been going around for, what, two years roughly now? 4 trillion in data center built outs over the next three years or something like that. And when you look at that, yeah, $20 billion in revenue does not sound like it's going to get you to enough financing to build those data centers to keep that model going.

So I will go through all of this to try and give us some perspective, especially with a little bit of seeming shakiness in the global economy and the US economy in particular. And I wanna remind everybody about how bubbles work, which is that during the elation period, everybody is focusing on what's really great. And don't get me wrong, AI is great, and for what it's worth, open AI is also great, although I don't think they're gonna win in the end, and people aren't thinking about the inevitable concern.

Problem doubts among the people who are gonna, in the end fund all this, which are the businesses shelling out money to use these foundational models. So great $20 billion in revenue, great pledging to build all these data centers. I don't know if that's so great, because they need to be efficient, they need to work well, and they just lead to more, I would say, fear, uncertainty and doubt ultimately.

But in the meantime, Altman stays famous, keeps making money, keeps hoovering up funds and resources. So that's my perspective. And I will say it's not just Altman, all of the open AI exec team, Hoover at the nonprofit there, um, are doing pretty well in terms of their own Comp nonprofit, and they continue.

Yeah. 4 trillion in pledged projects that they're involved in. And so a lot of people are looking at this and saying, Hey, the only way this is gonna work is the government's gotta back 'em out because maybe, maybe open AI is too big to fail, and we know how much the administration loves their AI players, especially US-based AI players.

And so who, who, who's sending them their snap cards, whether the government shut down or not, right? And, um, thi this is, you know, this is the question, and in, in the article that, that this is based on, you know, some folks from the government including the, uh, the, uh, the czar, the AI czar, whatever in the, in this space, said, look, there are four or five frontier models here in the us. We're not, wed to open ai, we're not wed to open ai.

However, if open AI didn't get the revenue that it's going to need to do what it's pledged to do, it will create a crater in, in the AI landscape, I think. But, um, you know, we've had craters on earth before. Kate, go Ahead.

Yeah, so you used the word however, right? Nice transition there. Um, in the article, they sp they wrote specifically, or, or, you know, talked about, um, backstop, they used the word backstop.

And that just sort of highlighted for me the entire article, not a bailout, not ownership, but for me, you know, it just makes me so uneasy when I hear about government involvement and ai, and as a Gen Xer who grew up in Cleveland, and this might be a, you know, this crazy parallel here, but I remember Mayor Dennis Kini, who refused to sell Cleveland's public owned, um, electric utility and, and people in Cleveland just went nuts. Uh, it was as if like the Browns made it to the Super Bowl or something. I mean, you know, it was that type of nut net Would be that thing.

But in all due seriousness, the moment taxpayer money or public ownership enters the picture, the conversation stops being about what's best, um, these long for best long-term and shifts to which side are you on. And that's what I remember. And so now when we're talking about taxpayer money and ai, it really concerns me that we leave the model, I mean, for the United States overall.

And we have always come out on top because we believe in, in, in, in fair, in this fair market, like what's best. And I don't like where I see this going at all. At the end of the day, it, it concerns me a great deal.

And yeah, I worry, I, I worry about this because I still believe that we don't see, I love ai, but I don't think we are looking at, at truly the right players yet. And so it scares me. Look, I, I will tell you this though, Altman's transparent.

He said all along, they're gonna need four to $6 trillion or some number like that to recognize all of his AI dreams or fantasies or nightmares or whatever you want to call 'em. Um, and, and, you know, because his, his vision of what open AI is, is much broader than just chat GBT. They wanna make their own chips and take Nvidia out of that equation.

They wanna own their own data centers. But why should we pay for that? Do you, I don't understand that.

Oh, a hundred percent, Kate. Uh, so listen, too big def you used the phrase too big to fail government backstop, whatever you want to call it. Um, I would say that a public utility in a metro, and by the way, I'm sorry, Stephen Foskey couldn't join us today, Kate, I didn't understand the Cleveland connection until just now.

I don't Say anything about Cleveland, But you know, there's, there's so, so, you know, public competition and public utilities, at least for distribution, that's, that's a tough, um, for on, just on an efficiency basis, a tough, a tough, uh, argument to make. But ai, I mean, come on, there's cohere there. Just in the private sector, there's, there's cohere, there's, there's Anthropic, Microsoft Building one, there's Gemini.

I mean, there's a, a, a, a Cajillion Frontier Foundation models out there. Maybe not that many, but plenty, plenty. There's gr Okay, so, so if, if there wouldn't be a crater, Alan, if OpenAI exploded and chat GPT disappeared, first of all, chat, GPT would not, or GPTI should say, should, would not disappear.

It would get, uh, uh, picked up by somebody else and used. So, no, there's the, the only thing that makes it too big to fail, hold on. The only thing that makes it too big to fail, and this is candy for you, that's why I interrupted you, is that under the current government paradigm here, if you grease the right palms and pay off and bribe the right people, then you become too big to fail.

And you get the, you get your backstop only because of the grift. I'm not gonna take that bait. 4 trillion pledged to other companies.

Those co I mean, Oracle, Oracle, um, stock went up by a third, a lot of it based on this huge open AI deal where they're gonna house a lot of the open AI infrastructure saying Google 38 billion. They, it, it's, it's cir it's circle jerk money. Yeah.

But it's playing the odds though. It's play the come out. We're talking about the stock market or the equities market.

It, it, that, that's the, you know, it's playing the odds. People will buy it up now and then in two years, if it looks like there's gonna be a regime change here in the US or whatever, or if there's the backlash starting, then they'll sell it off and short it and so on and so on. It's, it's, you know, it's sports betting.

It's sports betting. It is. And, and I think, you know, make no mistake, the under the present administration's model is they take a stake in these companies, not like the TARP model during the Great Recession, where the idea all along was to sell back the equity.

Totally different scenario. These Guys Wanted 20 different scenario. Yeah.

They wanna take a stake in it because they wanna play the market too. This is a casino, and they, they, you know, the government's gonna have its place at the crap table, right. You know, and, and pray.

We don't roll sevens, But at least the family that's, uh, that's, that's presiding over the government. Yes. And that's, that's the point.

I mean, guy, what you're saying, it, it's, it's awful what's happening. And, and, and winners are being picked with tax pair money. We're, we're freezing the market.

We're shrinking imagination, and well, we're Shrinking innovation. I believe this was the, the butt of my shimmy says last Thursday in an article I wrote about tech heal thyself. We, we need, we need a tech industry where two guys in a garage can change the world.

Totally. Not, not someone who gets a hundred billion dollars from the US government for a 10% stake. Absolutely.

Where, where, you know, ideology is replacing innovation, and that's gonna lead us down to bankruptcy as, So if you're watching this and you're wondering how this applies to you, because outside of where you're gonna invest money or buy stocks or whatever, you actually are a business, um, trying to figure out if there is an AI to bet on, or if, if your investment in open AI as a, you know, for, for, for company use or organizational use is good or bad or whatever. I just wanna say, don't get too elated. Don't, but also don't worry too much.

Just be happy. Use what you've got. It's gonna be far easier to, yes, it's gonna be far easier to switch models.

I mean, you can already do that with half the services that people use, like perplexity to what have you. It's, it's gonna be far easier to switch models. Um, I I, if OpenAI goes up in flames or whatever, then, uh, then you know, a lot of then like, you know, the cloud stickiness that we had to deal with and so forth, or I'd still have to deal with.

Agreed. Okay, you want to take the last word? Just, I think what is best is let's not, as guy just said, I hope that we can definitely, beyond any doubt, have a lot of choices at the end of the day, because that's what has always made us great, um, is having innovation, imagination innovation.

This is what we need today. So don't worry and be happy. You know, to paraphrase Jeff Goldblum in Jurassic Park, and I did it in my shimmy says, where he says, life will find a way, innovation will find a way too.

They'll figure it out. All right, let's take a break. We're gonna come back and talk.

Look, this AI craziness isn't just here in the us. Europe has it too. You're watching Textron Gay, You've earned it.

The spotlight, the responsibility, the weight of teams, companies, and entire industries fall on your shoulders, lives depend on your decisions, your home life included that work. You are protected physically and digitally. Nothing gets through your team without a fight.

But in a globally connected world, everyone sees you, including those who mean to cause you and your organization harm. And now home your sanctuary attackers see an opportunity. Your digital front door is wide open.

And what compromises your home can breach your boardroom. Because the devil's greatest trick isn't targeting your workplace firewall. It's convincing you that your personal life isn't at risk.

Black cloak, digital executive protection, defending the new attack surface your personal life. Hey everyone, thanks. We're back here in Textron Gang.

We're gonna go to our B block. It's actually sort of related to a block in that it's around ai, um, and infrastructure and booms and bust or potential bust and what have you. Um, you know, there's a report over on techstrong it that the AI data center building boom is not confined just to the us but it in fact, in the EU and the rest of Europe.

You know, they, they want, they're on the quest for, uh, it sovereignty, data center, sovereignty, AI sovereignty. And so they're in their own boom of building out data centers, AI factories, whatever you want to call 'em, um, and all of the requisite pieces that go with that power generation cooling, finding unique and novel places to build these things like above the Arctic circle, under the sea in space, I don't know, in the mouth of a volcano. Next, maybe I, I'm not quite sure, but, um, you know, this is Mount Doom, perhaps.

Sorry. Yes, exactly. Uh, this is, you know, this isn't just a US thing, the eus, you know, going through the same thing.

And I, I think all of the caveats that apply that we just spoke about, and potential cra and all of this, uh, Europe, Europe's faced with the same thing, maybe on a slightly smaller scale, but it's the same thing, guys. What do you think? Well, for me, I saw, um, block a more around legacy and block B, what we're talking about as far as them looking into the future.

I think having another cable going across, you know, the bottom of the ocean is, is good. Um, I think we were create, we're now, right now, we're not gonna, we're not gonna have a choke point. And I think by building more into Europe, um, we're creating resilience.

And I think that they're doing, Europe is doing things that I think a little bit different. It more like I would have seen in the United States where they're sort of looking at different technologies and they seem to be more open to not just this one. I and I, and hey, grok, you know this.

I like it. And I think that, you know, you see more players in Europe and, and I really like that. And, and I see more of a common sense, um, on how they're building it.

Like I said, the, the cable on, on the bottom, you know, they're getting rid of these possible choke points. And I think that's good. You're referring Kate to the Amazon, uh, a AWS, uh, 320 terabyte per se, terabit per second cable, they announced, uh, transatlantic.

It's not gonna come online for about three years. Um, But they're starting. There's no, uh, so, so it's like two stories.

Uh, one is this sovereignty idea, um, uh, building the building out of data centers of, of various kinds, neo clouds. There's, I mean, Europe is a hotbed of neo cloud activity, um, both in terms of innovation and diversity, uh, like a lot of different companies, but also in terms of activity. But it's not just neo clouds.

It's, uh, it's, it's a lot of, it's a lot of, uh, uh, data center buildouts. And that's, that's one thing. And then the other thing is this connectivity side symbolized by the Amazon cable.

Um, I wanna talk more about the first, I think that the, the, the continuing build out of the global network, um, even privately led, like with the, with Amazon is, is an inevitability. And it will never be fast enough. And never's a long word.

com bubble by laying so much fiber. I think that, uh, the data needs are, are so exponentially large right now that there's no way the global network can keep up. But just to say a word on by the preferred topic for me, which is this European build out, the, the, i i I, I don't like to use another phrase, uh, perfect storm, but there are just so many factors driving, um, European governments and European companies to do these buildouts.

Um, the, the, the network is one, but, um, the collapse not collapse. The, the, the crumbling of globalization, um, where it's lar it's becoming more and more each nation for itself, especially in terms of strategic technologies. And it's pretty clear that no one wants AI to become strategic and not owner control.

Some of it as well as information flows. And a desire to stay a little bit distant from the US orbit. Alan, I don't think that it's smaller.

I think that it's, if it's smaller now, it's just gonna get bigger. Because Europe is now the, the, the probably, uh, the better environment for global innovation, um, than the us which may only be temporary. I think it's only gonna be temporary.

But for the moment it's, you know, put, made everyone really mindful of being able to own their own stuff, to use a, a different S word than only came to Mind. No guy. You're right.

The sovereignty issue for Europe is huge because, I mean, the, the head of Microsoft France testified to the French government that even though they operate data centers in France, if the US government asked them to pull some information from a data center, from data that was hosted in a data center, oh Yes, that's right. They would do it. And that scared the hell outta the Europeans.

And, and so you have like my friend Andreas Prince, who's now the head of it, sovereignty, it's Sosa, where they're building European data centers staffed by Europeans on European soil and totally independent. But guys, I, I have an, another article I just read. I mentioned that innovation will find a way Shimmy says an article, but there's another article that I think will be out by the time this airs on Tuesday morning.

And that is around all your data centers are dead. 'cause here's, here's, here's the, and I, I, I, this isn't my idea. I got it from a guy named Guy Mallory.

Not you guy, a different guy who is with, uh, Comscope over in London. Uh, it was a LinkedIn post series he worked on. So here's the deal.

Your average data center is set up to handle, I think it's a 10 megawatt rack. Does that make sense to you, guy? I'm, I'm not an expert on that.

Yeah. Uh, the k uh, yeah. 10 megawatt.

Uh, 10 megawatts. No, not megawatt. Uh, wat kilowatt wat kilowatt?

10 kilowatt 10 kilowatts. Yeah. 10 kilowatts.

10 kilowatt, 10 to 15. Pretty average right now Is the rack, The specs, lemme just get ahead of you a little bit. The specs, like the top, top of the line specs right now is a single is a one megawatt rack.

Um, the most advanced specs that's being promoted by the open compute foundation, so the new rack led by Facebook. Yeah. The new the new open AI racks is, is i, is it megawatt or I think those are in the 2, 2 50 to 500 kilowatt range, which is just enormous by the way, the, the, I'm just referring.

So what if you say open ai, I think you mean open rack. I think you mean open rack, right? Right.

And no Nvidia, Nvidia, the Nvidia racks of the new Nvidia, they're also, so, yeah. Right. So, so the, the biggest racks getting put into production right now, um, just getting put into production right now, um, whether open or not are in the 200 to, to, to 400 kilowatt range, maybe a little higher, which is an enormous amount of power Size, rack size matters.

But, but here's the point. Whatever it is, 10 kilowatt megawatt, gigawatt, you know, Managed to the data center people. But go ahead.

No, it, it does. The, the important thing is it is exponentially, exponentially, way more than every data center we have now can handle it. We, we can't handle the power.

We can't handle the heat it throws off, and we don't have in place what we need to cool them down. So if you are just gonna run legacy payloads in your old data centers, that's okay, but how long are you just gonna run legacy crap without moving forward? And so if you're gonna move forward, these data centers are useless.

You can't, you know, you're talking about retooling them. You gotta bring in 10 x, 15 x, 20 x more power. You've gotta, it's a kin, this guy guy, uh, Mallory said it's akin to trying to cool a volcano down with a, with a, with a, a desktop fan or, um, you know, you just not, Right.

I, so I, I agree. Absolutely. Um, guy with what Alan is saying, because we continue to build on what we don't even know yet we're building, um, as if we're still thinking legacy, but we have no idea what the AI chips can actually really do yet.

And, and it's, we're building, you know, we're building yesterday's infrastructure for, and We dont have, right. Even things designed today may are probably gonna be obsolete by the time they're built It pr probably, yeah. You know, yeah.

You know, three years, like, I don't even think we can look three years. I think, I think we're talking like three months, six months, the speed at which we are able to innovate if we're allowed to truly innovate. I, I feel like the potential truly is exponential.

I, I feel like we have no idea yet of what we're actually looking at. I don't think that we need, uh, the fastest and the biggest, um, for innovation. I think innovation is kind of, I think I like, I like you quoting Jurassic Park.

Alan, uh, innovation finds a way, and part of innovation is, is the deep seek type innovation, which seems to get more from less. My theory about deep seek was that, uh, rather than using bigger systems and more data, they use one commodity that China has in abundance, which is people to, uh, do a fair amount of, of, um, uh, curation and, and massaging of, of the data used to train. And the white paper that was published to support what they did doesn't really contradict that.

It doesn't highlight it. So I could be completely wrong, but let me, let me, let me give you a perspective that, that, you know, from the data center manager and, and data center architect's perspective, I've talked to a fair number of customers about this over the last several months. And, um, they are, um, absolutely, they absolutely have their sights on, um, especially the bigger ones.

Uh, the ones providing service to other companies, um, whether they're service providers themselves or associations or in the, or in finance or those industries, um, and telco, where they need to, to, to provide service to wide at ranges of customers at scale. They're absolutely looking at denser higher power environments using direct, direct liquid cooling or even fancy stuff like immersion cooling, hybrid cooling. Um, and they are largely doing this in existing data centers, um, where they can increase the, the power availability, but mostly free up a lot of space for more standard systems that are now less expensive.

They can run their standard workloads. And let me remind everybody that 95 to 98% of the application estate in, in a typical enterprise, um, is not running an AI workload or not running one at scale. And the real place where this hu high density, um, and very high, uh, power draw racks and so forth is needed, is for running, um, uh, ru uh, in either training or for inferring inference for, uh, at a large scale for ai.

And those applications are relatively rare still. They just get, they just huge and amazing and incredible, and they have a really wide impact. So they get all the press.

So, um, color me extremely doubtful that, um, this is going to be a, a, a stifling of innovation around ai. I I'm not saying it's gonna be a stifling of innovation. I am saying, when you look at the $3 trillion that appears to be earmarked for AI factory and AI data center build outs over the next three years or whatever it is, you could understand that if, if, yes, today the applications are rare compared to legacy, but if you, if you buy into the, to the curve of, of AI adoption and, and on and disruption, 3 trillion may not be enough because, you know, what's our legacy data center footprint that may or may not be out obsolete?

And, and this is a lot like a moon mission where we have to, we're counting on inventing technologies that don't yet exist, right? And I'm not talking about tank, right? But, um, So what'd you say, 3 trillion over three years?

Something Like that? Something like that. Yes.

8. That's A huge amount of money. It is rough.

It's less than 1% per year of the, of the, uh, annual global economy. And it is going to be, it's shifting spend. Like, so overall it spending should go up the share of it towards AI training and large scale inference that all these data centers and these, these, you know, high density buildouts are needed for, is also gonna go up.

Overall impact, I would say minimal. I just wanna put it in perspective. I I, I'm obviously the doubter in this conversation.

I'm not, I'm not, uh, there, there, there absolutely will be an issue with capacity. There won't be enough high density If three trillion's not, not a big number. Do you?

Can you give me some? Yeah. 20 billion.

20 billion. I take, I take 20 billion right now. I'll even give you 10% of the company.

I, it's really, it's so hard to have perspective on. I agree with you. It's so hard to have perspective on these things these days because you know it, Hey, listen, a half a mil, you know, like if a half a mil fell off of the open AI truck, uh, Sam Altman wouldn't even get a fever.

He wouldn't even get a, a low grade fever. And I would be set, I would be so happy if I could just pick that up. But we do need to put it in, in perspective to understand its ultimate impact on us and the organizations where we work.

Yeah. I, I just look, I I think it is a giant gravity sink that's pulling money out. And part of it is There are minds share, Alan, our minds Share.

No, no, not the money, no guy. Yeah, there is minds share, but part of it is, if we buy into having to do this, and this goes back to my innovation, we'll find a way if we buy into having to do this to support our AI future versus finding better, cheaper, faster innovation, this stifles Absolutely. Alan, you're absolutely Right.

Instead of saying, let's just keep throwing more money on the bonfire, and we'll put some band books on there too, so we get a bigger flame, right? Instead of doing that, we should be saying, Hmm, you know, what did those Chinese guys do it? It's, there might be something to that.

There might, we should look at other things Context, right? Over, yeah. How can we use our existing data centers for this?

So let, let me, let me wind up my own thoughts on this by being much less negative, which is that I think both are true at the same time. I think that, um, there is no, there's really no way for capacity to meet demand over the next three years, just like you guys say, uh, there is no way to create enough, build enough high density compute, maybe in the three month window like you're saying, Kate, um, uh, organizations and users are gonna feel it. But that in turn is going, is what is going to drive the, uh, let's call it the Chinese innovation, um, that you were just alluding to, Alan, which is creative and interesting ways to get, um, real value out of less dense, let's say, out of, out of the capacity that is available in plain old, remember, plain old telephone service pots that, let's call it the plain old data center pot.

I get it. We get it. Hey, I'm at, we're out time for this one.

We got, we've still got another block we're gonna do. So let's take a break, okay? We're gonna come back to you and let's talk about something mundane like cybersecurity and ai.

You are watching Textron Gang Discover Textron Group, the epicenter of tech innovation. We are your go-to for reaching IT, leaders and practitioners worldwide. Our secret impactful content that sparks awareness, engagement, and top quality leads with us.

You'll access editorial websites, streaming videos, virtual events, custom content analyst research, and more. Join our satisfied clients. Let's revolutionize your tech journey.

Contact us today and tell your story to the world in the most powerful way with Textron Group. Hey everyone, we're back here for our third and final segment on today's gang. You know, there's a recent story up over on Security Boulevard about a survey from S code, uh, surveying executives on, I don't know if I'd call it paranoia, but maybe their fears around, um, you know, the risk tied to AI generated code and whether that AI generated code is, you know, totally AI generated, or let's call it human assisted.

You know, we're turning out more code than we ever did, and we're turning out some, say, more vulnerabilities than we ever did. Kate, uh, why don't you kick us off on this? What do you think?

Yeah, so just, um, highlight more vulnerabilities. So 2024, we saw 40,000 software vulnerabilities. That's the highest, highest that we've seen as, as we have tracked.

So talk about, you know, trending and data. So, um, 400 cybersecurity leaders in the US and UK revealed 100% of the firms have begun using AI generated code in their code bases. 65% of these respondents say that they've seen an increase in vulnerabilities since doing so.

And so the data point here, which is striking, is that while AI boost pro, uh, productivity, quality and speed, but, um, there's no centralized governance around this, around AI adoption. So from a cybersecurity point of view, that's really, really scary. 0.

0, um, new, you know, revolution here. And what I will say is that LLMs can be and are being used as new attack vectors. So while mass AI co-generation with sparse governance, we are creating this hidden technical debt that is threatening resilience and trust.

So while I look at the blocks that we cover today with Block A being around, um, anchoring legacy, and Block B was about building foundational infrastructure, you know, it's nice to highlight and it's unfortunate, but it's typical that cybersecurity comes in at the end. It, it kills me, it just kills me. Um, but you know, this is our warning light, uh, that even the best infrastructure and funding won't matter if, if what we're building is structurally weak and ungoverned.

So what do you guys think? Alan? Guy, Guy, you wanna go first?

You want me to jump in here? I have some thoughts. Well, uh, let me start by being that guy and guy who points out that while this is a great survey, it's, uh, I think it was conducted either by IDG or ID sibo, perfectly reputable for, um, research firms able to conduct great, uh, market research.

And it had 400 respondents. Um, they weren't all leaders. They're described as CISOs and security practitioners.

I think that actually makes it better, um, if you're understanding people who are, as, you know, both on the strategic side of things as well as the tactical day-to-day side of things. Um, and that's encouraging. Um, what what really struck me, um, out of the, uh, survey results, um, was, uh, that eight out of 10 of the respondents, um, say that their organizations lack visibility into how AI is being used in app dev.

There are easily 10 ways to use AI in app dev. It's not all about just coding. And I guess if, if there are 10 ways, then there's probably a hundred vulnerabilities of being introduced, um, or potentially being introduced just on a broad ca category basis.

But the visibility, I think it starts with that lack of visibility. Very much so. I don't think it's quite possible, you know, um, in most organizations to tamp, uh, AI based code generation down by individual developers using whatever the fricking tool they wanna use.

Um, I think there's a great push right now towards integrating, um, uh, preferred and guard railed AI into code development, if we're just gonna talk about that to encourage developers, uh, to use that because it's easier for them. 'cause it's part of the IDE or the Yeah. Or part of the workflow.

Um, but eight outta 10, that's just, that's just sounds like A Crest commercial. Well, except it's like, except in the inverse. Instead of eight outta 10 dentists recommend it's eight outta 10 security professionals are saying, stop.

We don't know what's going on. Well, so I I, I have some thoughts here. First of all, I find it amazing that 100% of respondents say their organizations are using AI in generating code.

Even if that number's wrong, and it's only 95%, that is just phenomenal. You know, I I, I've got an article out yesterday and today or Monday, excuse me, and I posted a LinkedIn article Sunday on it about AI use in journalism and media. And, and right in the second line, I think I said, any media or journalist company who tells you they're not using AI is lying.

They are, It, this points to the quality in my, in my opinion, points to the quality of the survey. Not just the eight outta 10, but, um, because anybody who, any security professional who is not saying that AI is being used to their company to code is just Outta touch. Right?

Outta touch. Yeah. Not, not the kind of person you wanna serve, Sleeping, sleeping at the wheel.

But, but here's the thing. As counterintuitive as this may sound, the only thing that's gonna save us from this is ai. No.

And, and I agree. I I, I think that we need to understand that we're all using it. We're, and we need to embrace this and start to, you know, we really desperately need to embrace it and secure it right now, you know?

Yep. We Continue. I mean, because here's the thing, Kate, what you said is a hundred percent correct.

Security's always the caboose. Yeah. Always.

Security's always the afterthought. It's like, and they're not gonna stop, they're not gonna stop using AI to generate code. 'cause the security guy is holding his breath and turning red.

Right? Or, or the security gal is stomping her feet and having a hissy. Yeah.

Yeah. Um, it's not happening station. It's like bringing own bike.

Yeah. It's like BYD, right? It's not, it's, it's like trains left the station.

Yeah. So instead we've gotta turn AI into our friend. We've gotta turn AI 100% into He is our friend.

Yeah. He is our friend already. Alan.

Well, he's my friend. It's my friend. I, I, you know, it's funny, I think of AI as a guy and not a gal.

I wonder why that is. Like, my car is a, a female. My, but I always think of ai.

Do you guys, what do you think, Kate? Do you think of or do you AI doesn't have You with all of my, um, it's funny that you should mention that because I thought with all of my AI as assistants, I actually name them per the device. Like my little shark robot that vacuums, I actually just call it vacuum.

Like I don't personify ai. You don't genderize or anything. I don't, I don't, I don't know why.

God, what about you? Well, I think generated ai, there's no doubt that it's a white man because the data used to train it. Yeah.

The data sets used to train it we're biased towards white male perspective. Fair enough. Fair enough.

But I, it, I'm, I I'm, I actually, that is a serious point as well. So let not just toss that off as a joke. It's a serious point as well.

Mm-hmm. It is, it is. It sort Goes.

But we are understanding that, right? I mean, yeah, we do understand that there's bias. We've been trying to deal with bias.

Um, We don't understand that, Kate, you and I and Alan understand it, and a certain elite of the, the people watching this stuff understand it. But I, I'm sorry. It, it, as in as a general rule, no, I don't think, I don't think people understand the bias.

People, users. Well, maybe we start to add that into our security config. Honestly.

Should do, Should do. Yeah. Absolutely.

But, but, but here's the thing. I do believe AI can help us help itself to generate more secure code. I agree.

To find vulnerabilities and fix them. 'cause it's the only way we're gonna be able to do that at scale. And even within GitHub.

I, I mean, wouldn't it be, you know, one of the problems that we have with the 40,000 vulnerabilities that I mentioned is that we keep reusing code that we know is vulnerable. When do we start to get rid of and really seriously de bloat legacy code? I mean, come on.

It's time people. It's time. Agreed, agreed, agreed.

I I, I think that, so I'm not saying their paranoia is not valid. It is. But but, but we always have to name it first, right?

We always have to name whatever. I, if I hear visibility one more time though, I mean that also, I'm really tired of the visibility story. This whole single pane of glass, this unicorn that we're chasing, you know, cybersecurity folks.

Yeah. It, it Can, can I give you a little corrective, Kate to, to, and I didn't make it up. It's very old to single pane of glass.

'cause I, I, I get an allergic reaction to that too. Um, the phrase I like, which I think Forester originated about 20 years ago, is preferred pane of glass or preferred workspace. Because if you prefer to have 12 what you want, maybe you have 12 modes in where you work.

And when you're in mode, what you, you want lots of data to be running across all those different areas. It's true in observability visibility, threat detection, that sort of thing as well. I think you, you, you, you want to be able to use the pane of glass that suits you best, even if the data there is also in two other panes of glass that you use.

So I think preferred pane of glass is fair enough. Fair enough. And when, when you walk into soc, right?

I, I mean that's what you have are all these panes of Absolutely. Of Data box myself. But We try anyway.

Hey guys, we're overtime looking at my clock here. I see we ran over. Imagine that.

Just the three of us. We could probably do this all day, but I gotta pull the plug. I gotta run over to CubeCon and, and, uh, get busy there guy.

I'll see you in a little bit. I hope. Kate, thank you so much for joining.

Appreciate it. Thank you for watching. Uh, we are gonna be live from, uh, uh, CubeCon today.

So check that out. We'll be streaming live later in the day, and then we'll be doing text drug again the rest of the week from CubeCon. And if I may say also, we are already streaming live on LinkedIn right now for Tech Field Day exclusive at CubeCon, which, uh, I am also participating in, although, um, I'm gonna be changing shirts before I go back there.

Very cool. All right, until then, Kate Guy, thank you. Thank you for watching.

This is Allen Hummel. We're out.