Nikesh Arora | All-In Summit 2024

Well, our next speaker is actually fortunate enough to have had seen his brand name turn into a verb.

One of the probably most prolific executives in this current generation is Nikesh Arora.

The big daddy of the cybersecurity space. These guys are really at the forefront of the industry.

There are very few people who consistently, time and time again, find a way to just persevere, be relevant. Nikesh is one of those people.

This is a man who has tremendous insight into technology. He helped turn Google into the dominant player in search.

This is the most innovative industry We're constantly paranoid from an innovation perspective. I've got to be on my toes because once we figure out how they did it last time, they're trying a new way to do it next time. This is the country where your dreams come true. And if you go around the world and you ask young people where do they want to go, they still want to come to America. I think this is one of the most successful democracies in the world. This is where capitalism thrives.

All right.

Ladies and gentlemen, Nikesh Arora.

My guy. Appreciate you. Thanks for coming.

David.

David.

How are you?

Sup, Bosco.

Let me just do this intro properly. Look at all the phones go up. Wow. You joined Google in 2004, although there was a nice prolific buildup to that career, but you joined in 2004. You left in 2014. You started in ad sales and you left as the SVP and Chief Business Officer of Google. Revenue went from $3 billion. I checked this actually just to make sure because it's staggering, to $66 billion when you left. And then you— because we're going to talk about that. And then you got seduced to go work with Masayoshi-san at SoftBank where you were vice chairman and president.

Yep.

That must have been interesting. Jason has— look at Jason.

He's licking his chops.

Jason has so many good questions. There we go.

But then you left.

Yes.

And look, I've known you for a long time. We've been very good friends for a long time. I was surprised because I got, you know, you called and you're like, hey, I'm going to be chairman and CEO of Palo Alto Networks. And I had known what it was, but I didn't really understand. And then meanwhile, in the last, what's it been, 7 years?

6 and a half.

6 and a half years. Market cap is up by 5x. You took a $20 billion company. It's $110 billion as it stands. I think you've tripled revenue. So this is clearly no longer luck. So now you're in the skill camp.

Oh, good. I'm in founder mode?

No, founder mode is cocaine.

I was wondering when that was going to start.

Wait.

I was wondering when those—

Do you have founder mode on you?

Because you're here now. No, he has to fly to Europe. You cannot bring founder mode.

They have no waffles on the plane. No waffles on the plane.

You can get founder mode in Europe, though.

I've heard. Let's just start and just— Actually, let's just start there.

Okay.

Okay. Um, you've seen a lot of different executives, you've played a lot of different roles, you've seen founders, you've advised a lot of founders. Tell us what, what, what takes— what, what does it take to be successful? And you can use these labels or not— founder mode, manager mode, whatever it is— but what does it take to figure things out consistently?

Look, um, I think you already put that out there, didn't you? Didn't you say that, uh, if you think about building great businesses. At the center of great business is great products. If you don't have a great product, you're not going to build a great business for the long term. And this is something— and I know Sergey's here— I learned that with Larry and Sergey at Google, that they were obsessed about product on a constant basis. So when I came to my job, I said the first thing I'm going to focus on is build a great product. But I think it's slightly different in consumer and enterprise. In consumer, you build a great product, you find the flywheel, you try and figure out how the flywheel continues to work. In enterprise, eventually you take a great product, you got to figure out how to get it out to all the amazing customers out there. So I think it requires a tremendous amount of focus, tremendous amount of sort of detailed inspection. But I think at the same time, you got to find a way of taking lots of amazing people, getting them on the same train, and getting them to execute at scale. It's impossible for one human being to do that at scale. So you have to have a lot of people doing it amazingly well.

Can you tell us—

That's the trick.

Tell us about that first story or that version of that story inside of Google, because you were there for a long time and a lot of good things happened. What was that like? What did you learn?

Look, Google has one of the best flywheels there is in the consumer space, right? So we were blessed that we were working with a product that nobody had ever seen, everybody wanted to use. It's funny, like every one of us worries about customer support. You didn't need it. It was an amazing, simple product, easy to use, free. And our job was to go monetize advertising. So part of that was how do you scale that around the world in every country where there's a single product with a single use case where you have to see how you can attract lots of advertisers. And that requires building a system. How you get thousands of people around the world to build a system and execute. So you build a system, you build a programmatic system, you look at stuff, you inspect, and you hire really amazing people who go out and do their best that they can.

And when you're doing that and the thing is growing so fast, what has to happen for you to go from running Europe, I think is how you started, to being the head of business there? What does that take?

Well, it was such an amazing juggernaut that you had to figure out how to differentiate. And what is interesting is when I joined, Google was 24% of global revenue. When I moved to the US, it was 49%.

You mean Europe?

Yes. And this is one of the few tech companies in the world whose European revenue was higher than the US revenue for a brief period of time. So I think somebody noticed.

And what happens? You get the call and you're like, we need you to move to America?

Yeah, I was on a trip to Russia trying to open an office there, which had to be shut down at some point in time for a bit. And I got a call from Eric Schmidt and he said, your boss is retiring. We'd like you to come here and do what he does. That was it.

And so why— what then motivates you to leave a job like that? Because you're kind of then at the top of the pinnacle. You see everything, you're meeting everybody.

I guess I wanted more. I wanted to do more, get involved sort of the overall business. I wanted to do some product work as well. I didn't have product jobs at Google. I was seen as a sales guy. At Palo Alto, all I do is product for the first 6 years of my life. So to be able to go out and do that differently, but I had to take a brief sojourn to my Japanese trip. Lots of good sushi and lots of—

It was a great vacation. Let's talk about it.

Oh, come on. It wasn't a vacation.

Well, a great surge, but I mean, this was the largest venture fund ever created, $100 billion. And you have this mercurial, brilliant, individual, Masayoshi-san, and he starts placing bets in a way that we've never seen. What was the genius in that and what was the Achilles heel?

Look, Masa is one of those people whose risk appetite grew as he grew older.

And you mentioned that because that is a very unique thing.

Yeah, because think about it. I have 2 young kids and every time And I'm constantly trying to de-risk them. I'm saying, hey, be careful when you cross the road. Be careful when you do this. You get married, people tell you, be careful, buy a house, go settle down. So we're constantly de-risking our lives as we get older on a constant basis. All of us do it. We don't realize we do it, right? Masa's the opposite. The older he gets, like, come on, let's go all in. He's like you guys, right? He wants to go all in. So he's like, Nikesh, I have a great idea. We'll put a billion, we'll borrow $19 billion. How does that work again? All you got is a billion. Yeah, I have one great idea. A billion in, $19 billion. That's what he did. That's how he built SoftBank Japan. I think he was the richest man in the world for 88 days in the last internet boom. Then he was left with a billion dollars.

Unbelievable.

And he built from scratch again. So yes, unbelievable.

So there were a series of incredible bets.

Yes.

Maybe walk us through some of those bets because there was Nvidia, there was—

So all those happened after I left, but that's—

Right. And then you had—

did Arm happen after you left?

That was where we kind of, you know—

Unpack it.

Sorry?

Unpack that. Unpack this.

Well, Masa likes a trillion.

He likes the number 1 trillion.

Yes.

Okay.

It's a good number. It's better than a billion. Beats a billion.

It is greater than a billion.

Yeah, it's greater than a billion, last I checked, yes. And when I met him the first time after we'd done a deal at Google, he was— we met when he came to see Larry, Sergey, and Eric and said, I'd like to do a search deal with you. I have Yahoo Japan. I tried to explain to him there's something called— you can't have 2 search engines both powered by Google in Japan. And to his credit, he said, you can, as long as the advertising systems are different. So if you look in Japan today, Yahoo Japan is powered by Google and Google's powered by Google, but the advertising systems are different, hence it's non-competitive. So he got that done. And then he says to me, showed me this plan. He was going to buy a lot of companies in telecom and get to $100 billion in EBITDA, which at 10 times EBITDA would be $1 trillion.

Okay.

So then that kind of fizzled out. He lost interest after a while. And then his next idea was to raise, I think it was 1, 2, 3, 4. $100, $200, $300, and $400 billion. Yes, for the Vision Fund. So that'd make a trillion.

Hold on a second. Let's— we're doing that. Yeah, that's a trillion dollars.

1, 2, 3, 4.

Yeah.

Yeah. You know, it's interesting.

And then he had this whole portfolio of companies, and there's a bunch of Japanese analysts who'd sit in the office, MBAs from University of Tokyo. And eventually they took all the business we had and forecast their free cash flow at the end where the DCF was a trillion.

So he was good at setting goals.

So he thought Arm was going to be a trillion-dollar company.

Got it. We were—

sorry, let me ask you, do you think was the mistake— not the mistake— is it about being financially oriented as opposed to product or impact oriented? Is there an orientation thing there where if money is the goal, it becomes a lot harder to achieve versus—

Well, look, money is a way to keep track.

Yeah.

It's not the goal. That was the way he kept track. I get it. But he was not financially oriented as much as he went by his gut. It's like many founders. When he believed in it, he was all in. He totally believed in it, and sometimes to a fault. And you saw that. One thing I did learn from, which is very fascinating, is like a person who's used to wanting to get things right, I'd make an investment with him. And then it'll be one investment, say, oh shit, that's not going right. Let me go and talk to the company, help them, help them fix it. We can get them up and running. He calls me aside one day, he says, Nikesh, you're spending too much time with the mistake. He said, if you go spend that time with a company that's growing at 3 times, they grow at 6 times. We'll make our money up 6 times in that company instead of you trying to fix that from half back to 1. So it's kind of interesting.

That's an incredible lesson, actually.

That's a phenomenal lesson for all investors.

Cutting bait, sunken cost. I mean, there's a million ways to say it. But you have to let your winners ride. You got to focus on the winners.

Go all in on the winners.

Go all in on the winners.

It's hard to do. It's hard to say, oh my God, I made a mistake. You go and say, I can fix it. I'm good. I'm going to salvage it.

Yeah.

That's an ego problem. That's a hubris problem. I don't want to have a mistake on my record.

Or a good person wants to help the founder realize their vision. And the cutthroat nature of this with the power law is such that 6x-ing something that was at 3x is much, much more likely than getting a 0 to a 1.

Do you apply that principle as an operator? And if so, like how at Palo Alto Networks?

Look, in the last 6 and a half years, I bought 19 companies.

Right.

We can show the— you can show the slides after this.

Oh yeah, yeah. So we got some great slides.

We got some slides that just—

Wow. Here's your stock. Good job.

Thank you.

Not bad.

All right, go back.

Revenue's pretty good.

If you go back, it's really been up.

What's the market cap now?

$110 billion.

$110 billion.

And when you started, it was at?

$20 billion.

$20 billion.

What you're seeing here is the economic principle of founder mode.

And the revenue and the operating income.

Yum, yum.

So how do you look at something like this when you were first approach for the job, how do you underwrite the job? What are you looking at? You said you spent the last 6.5 years in product. Did you see a product that just— was it missing something?

That's the thing I want to hear about. Did you go all in on one or two big things? Yeah, was it broken? That was always the Steve Jobs model, was pick the winner and go all in on it, get rid of all the other stuff. Does that principle apply here or is it—

No, this is a different principle. Look, it's a $180 billion industry on an annual basis. The largest market share was 1.5%, which is us.

Hmm.

And it's a subsector of technology with the most amount of fragmentation. And look around, you know, Benioff, I think, is gonna be here, builds a platform for Salesforce. You have ServiceNow, you have Workday. There is no cybersecurity platform. You sit there and say, this is a phenomenal opportunity, one. Two, it's a company that's fully public, so I don't have to deal with voting controls and founders who I have to deal with, which have different motivations. It's an evergreen sector. The more we get connected, the more people want to hack. The more you're going to connect it, the more data is there for people to take away. So you're not going to have a demand problem.

Yeah, sadly.

If you can go into a sector where there's no demand problem, you can look at it and say, what did everybody get wrong? Say, well, everybody sort of lived in their swim lane. So we were in our swim lane. We did one thing. There are 5 swim lanes in cybersecurity. In 6 years, we looked forward and said, where is the world going to? It's going to the cloud. there's a bunch of AI. That was our sort of plastics moment, cloud and AI. So we said, let's not go reinvent the past. So one of the things I also learned during, you know, my time at Google and NASA is like a lot of people get hung up in trying to make the stuff work, assuming everything around you is going to stay the same. So saying that we're just going to focus at, assume that 50% of the world is in the public cloud, what's security going to look like then? Assume latency is low, you can process in the cloud, Data storage is cheap. What's going to change? So we built for that. We bought 19 companies. We went and looked at how everybody does M&A and who failed.

Oh, what did you learn from that?

Well, we learned that things trade at a price for a reason. So very often people say, you know what? Ah, number 1 is $1 billion. Number 4 is $200 million. I can take $200 million and clean it up and fix it. It's at $200 for a reason. And the billions are billions for a reason. They'll still be around.

Yes.

So why don't we buy the guy who's worth $1 billion? We'll be number one. We'll be leading the market. We have brute force and go-to-market. We'll go use that. And we're probably gonna slow them down a little bit because they're a larger company. So we'll compensate for slowdown with go-to-market that we bring to them and we'll let 'em lose. So we're the only company where when we acquired companies, there's a funny story. So I got a guy who says, oh great, we're buying a company in cloud security. I'm the senior vice president of blockchain cloud And AI. I'm like, great, welcome to your new boss. He's like, what do you mean? I said, that's the guy you're going to work for. He's like, we just bought his company. I said, yeah, he kicked your ass with low resources out there in the market. You're going to learn something from him.

There you go.

Welcome to your new boss.

Well, you know, there is an analogy. You share a passion for basketball as well. I see you all the time at the Warriors game. I don't think this is a jump to say watching that team play and how they manage talent and play as a team definitely informed how you play the game. Yeah?

That too. Not lately, but yes, in the past.

Yeah.

But look, we've done that 19 times. We had 7 out of 10 we've gotten right. We still possibly have the most number of founders who still work for Palo Alto.

Yeah, actually, can you explain that? When you buy a company, isn't What's the typical motivation? Wait till I cliff, it's a year, and then most people just vamoose, they're gone.

No, no. So here's how it works. If you come to Palo Alto, we'll take your equity away first. We'll say, I'm going to give you back 1.5 times your equity if you stay with me for 3 years.

Oh, wow.

To the founder? Yes.

Okay, wait a second. So the founder owns, let's say it's a $100 million— a billion-dollar company. They own 20%, they got $200 million. You say, hey, stay again, I'll give you $300 million.

Going to work for me for 3 years.

Pretty good deal.

Because when you buy a company, you're buying half a product and a full vision.

Right.

I lose the vision part of it, I get half a product.

Right.

How much of the success is predicated on the engine at Palo Alto Networks to drive sales, to sell into the enterprise? How much do you come in and then the founder feels like there's an interference model now that's like, how are you getting in my way? And how do you manage that balance?

So what happens is like, look, the customers and security want the best product. That's why everybody lives in their swim lanes. So we said we gotta be in multiple swim lanes. To be in multiple swim lanes with multiple people saying, I've got great products. You know, in enterprise there's this bizarre thing called magic quadrants, which Gartner has. And your badge of honor is you're in the top right, which is the leader's quadrant in Gartner. When I joined Palo Alto, we were in 2. We're in 24 right now. in the top right. So now when you go to customers saying, hey, I got some great products, you're going to sell me some good and some bad. I'm like, actually, no, you pick. There's all 24 are in the top right and they all work together better. So for that, we need the founders building the product and staying there. Yeah, they do feel a little— sometimes they feel like they're being directed. But there's also another rule. I had a wonderful conversation with the founder, which was my first acquisition, and I think we didn't set the bit right, so we had to fix it in future deals. Paid the founder a reasonable amount of money, probably $850 million to the company. He had about $150 million, he was going to get $200 million. Then he comes into my office and says, Nikesh, I got a problem. Um, I think we should do it this way. You're telling us to do it this way.

Hmm.

I'll say, because this way is the way it's going to work for us. He's like, yeah, but when I came here, you know, with my company— I said, oh, wait a minute. I said, have you ever sold a house? He's like, yeah. I said, who decides what— and you get to stay in it— who decides what color the wall is going to be painted?

The new owner.

He said, the new owner, of course. I said, thank you. So he stayed there. He really liked us, and he stayed for 3 years. He made 2.5 times that money he got. But from then on, we changed the game. When we acquire a company, we sit the founder down and say, okay, the lawyers will do their thing. you're going to sit with my head of product and design a product strategy we both agree on.

Yeah.

So we don't buy a company until we have a joint agreed product strategy with the founder.

Now in Cisco, Oracle, Salesforce, they've all kind of had this M&A playbook that they claim is part of their engine of success. How differentiated is it for you? What do you— what's kind of the biggest contrast for your playbook versus those engines?

So biggest contrast for us is We like to buy a product which we can integrate and sell to customers. We have a go-to-market engine. We like to keep it the way it is. I think if I'm gonna buy a company at 8 to 10 times revenue, I'm just overpaying for customers and sales. I have all the customers already.

Right.

Why would I pay 8 to 10 times revenue to buy a customer? I already have it on a different product. So I'd rather buy the product, use my go-to-market capabilities, go sell them to the customer base, Unless I can take the 2 companies, merge them, and I can make it worth 16 times. If you as an investor can buy both our companies and enjoy yourself, why would I have to pay a premium to buy it at 8 times revenue? So we're very clear, we don't want to buy customer bases. We want to buy products which we can integrate and sell into our customer base.

Let's talk a little bit about the threats that are out there in the modern world, how they're evolving with artificial intelligence. Obviously, it can be used on both sides of the of this competition to see who can protect information and then who can steal it. Who are the actors? What's the motivation today? And how are they coordinating? Because it feels like there is now this new allegiance between American hackers, very young, anonymous, working with— to do the social engineering, working with some brute force tools out of China, Russia, other places. Who's orchestrating these very large hacks that occurred at the casinos recently? And then we can get into how should our government, if at all, be thinking about stopping these and partnering with corporate America to neutralize these threats? Because some of them are involving the governments of these countries.

Yeah.

So I think, look, if you trace the history of cyberhacking, You had these big hacks which used to take 30 or 50 days to figure out. People were doing them as a hobby. You'd think of your notion of a hacker was some kid sitting in their parents' basement who didn't get out of there on his little PC trying to hack this and trying to get all the data out of there. And then suddenly people discovered, wait, I can get better than this, right? Because it was usually to prove, it's a badge of honor. Oh, I hacked into this database or I hacked in there, I got in there. You guys aren't strong enough. Now, as the world got more connected, What happened was people said, wait, why am I wasting my time hacking one user, one company at a time? Let me go after a piece of supply chain. If I hack the Exchange server, everybody who uses an Exchange server is fair game. If I hack an agent or Norton antivirus, I can get into everybody's computer. If I hack a large email provider, I can have access to every dissident's email.

Hmm.

Which is when nation states got involved. Nation states said, wait a minute, if I want data, why bother hacking one person? Let me go hack the backend and get in.

Let me hack Gmail.

Yes. More effective, right? So when that began to happen, nation states started getting more— saying, that's interesting. If I can do that, I can destabilize nations. I can get data about other people that I want. So that became a bit of a nation-state activity. Now, cybersecurity offense is way easier than defense. Defense, you've got to be right 100% of the time. Offense, you're going to find one door. So, so put that aside. Then what happened on top of that is that nation-states started cultivating these entrepreneurs in the hacking world, saying, listen, that's how you keep your skills up to date, is you go after stuff. We'll look the other way while you're going and doing it, because if we need you, we would've found a way. And then we discovered this notion, wait, there's tremendous economic value now in hacking. So it became ransomware. People wanna say, And there's like a magic number. They ask for $30 million or less because that's director's liability insurance.

Wait, sorry, sorry, sorry. When you get hacked for ransomware, $30 million is what?

Is what companies can give you where it's covered by insurance.

Oh, it's covered by insurance.

Beyond that—

So it's purely working backwards from that policy.

So there's about $2 billion that's been paid in the last 12 months in ransomware.

Wow.

Wow.

Now, if you think— here's the anatomy of a hack, right? Somebody says, I found a SolarWinds server, it's hackable. So they, some set of guys go quickly and plant themselves at 1,800 SolarWinds servers, which are exposed to the internet. Then there's a separate industry subsegment they sell it to saying, listen, I'm only in the seating business. You can go run ransomware as a service negotiations with customers. Wow.

That's their go-to-market.

Yes, this is their go-to-market amplification to systems integration. Yes.

Yeah.

And then there's a third set of people who are payment clearing people say, I'll collect the money. I know how to process $30 million a day.

Oh my God.

That's right. This is so sophisticated.

And where are they geographically? Is it all over?

Everywhere. Everywhere where extradition treaties are light.

Are there specific foreign nationals that move to jurisdictions to do this? Is this like—

There's a lot of people in the world out there who do this, and they're hard to find. And remember, think about the enforcement. Where are you going to go? Go to your local police station?

You're freaking me out.

Somebody takes $1 million away from your bank account, who are you going to go to? Local police? The guy says, actually, sir, this looks like somebody in Greece.

Yeah.

Do you know our, like, panic attack?

Yeah.

Let me shift the conversation. I want to talk about AI for a second because you mentioned it as well. But I want to first ask it to you more as just a smart observer of the market. You're in the market. You've invested in a bunch of companies as well. What's the state of AI? Take it however you want.

Yes. I mean, look, what's interesting is I think a lot of people are chasing LLMs. I think there's a very well-established expectation out there these LLMs will get smarter and smarter, inference will come, latency will go down, uh, cost to deploy, cost to train will all come down. So the good news is we've seemed to have established a nicely competitive space out there between all these people that you can expect some sort of economic rationality to prevail. And a lot of people are sort of investing a lot of dollars to get it there. And thanks to Mark Zuckerberg throwing out open source models, he keeps them honest and fair everywhere around there. So we'll all get access to these models. But for the most part, as you get into the application of these models, I think the world changes in consumer enterprise. In consumer, you got to figure out how these models are going to translate into consumer services and make them better. And you can see that the question is, do we get a whole new Google that's formed or a whole new Facebook that's created, or do the existing players move fast enough to embody, sort of to embed AI in there, and our hooks those services have into us are so strong that we don't shift our usage?

Yeah.

So let's park that for a second. We'll go back there in a minute. On the enterprise side, it's not useful unless you can train on my data. And you'll discover 90% of companies have bad data. 90% of companies. Like, how did you solve this problem? I don't know. I don't know how I fixed the last firewall that broke down. If I don't have that data and if I don't have 10 good instances, how do I make it work in the 11th instance? So we're all busy refactoring our data, figuring out how to collect good data on the enterprise side, which is going to happen. It's all the easy stuff. Sebastian will tell you, I've got it figured out. I'm going to answer questions. Those are easy How much balance do I owe you? When do I owe you? Can I pay you tomorrow? No, you have to pay me today. Even an AI bot can answer that question.

Right.

But it's very hard to say, my firewall broke down, I don't know what happened, how do I fix it? 'Cause I'd need a lot of data. So I think on enterprise side, a lot of companies would have to do a lot of work to get their data sorted. And that's in process. What we've done is stimulated all of us to go out and get that figured out. On the consumer side, it's gonna be very interesting. I think we can all imagine a future which says, Hey, my favorite phone or favorite hardware device or favorite interface, go book me a ticket to Geneva, which is where I'm going after this, and book me a restaurant and a hotel room. Now you just in your brain said, wait, wait, wait, wait, I just did Booking.com, I did OpenTable, and I did Hotels.com. Now we're gonna see this happen. Who's gonna control the user interface and whose agent is gonna talk to who? Right. Try telling any of the existing app guys that, listen, suppress your UI. I'm just going to send you an API call, send it back to me. I'll control the data about the consumer.

Yeah.

Let's see how far that lasts.

I mean, that's— yeah, you're just handing over your business to them. Yeah.

Well, then what's going to happen? One of two things happen that always happens, right? These people become the legacy players, and you'll have new companies that are formed which are agent-based only. It's like, you know, half your fortune is better than none. So if I started a company tomorrow, I would say, listen, I only have an agent that does airline bookings. Just pay me 20%, I'm good. I don't need a brand. What happens then? So I think there's going to be much more upheaval in the consumer space than any one of us realizes. I think 5 million apps will be redesigned in the next 10 years. They'll all become agents.

Right. The ones that want to survive will do that first.

But it's very hard.

Yeah.

Your margin's my opportunity, I guess, is the way we say it in the industry.

Yes, yes. But it's very hard. Can you try going to any of these large branded apps that sit on everybody's phone and say, listen, shut it off.

Yeah, become a service provider of data.

They would, they would just talk to Siri, just talk to whatever, whatever it is.

Yeah.

And we'll get it done for you. AI and attacks and sophistication of the attacks that occur How often does human factors, phishing, tricking people, come into play these days? And how much of that is going to be exacerbated by AI, deepfakes, et cetera?

Oh, the attacks are most simple. Most simple.

Explain.

You know, we had a whole pen test company, and they do this thing. They said, listen, we're going to penetrate your defenses. It's not, not possible. It's just impossible. We're gonna figure it out. The guy goes in the morning, 8 o'clock, to the parking lot, drops a bunch of USB sticks with little tape on it that says, my home videos.

Oh my God.

Wow.

He drops about 25 of them. 6 of them log in with the USB stick in the computer in the office. They're in.

Oh my God. Oh fuck.

Oh my God.

So great.

So I don't know if you need to go like get Battering ram and break your door down. This is human behavior. Human behavior. And they possibly said something more colorful than my home videos on that, but I'm just going to keep it PG here. So you can decide at what point in time your curiosity gets the better of you.

Mind you, it's with a Z. That got 100%.

Yeah. So it's like, these are not hard attacks. We sent an email out to everyone saying, National Pet Day. Please take a picture of your fluffy pet at home and upload it to this website, and the person who does it will give $10,000 to the SPCA. Oh my God, you should have seen the beautiful fuzzy pictures uploaded to this hacking site where we had all your details.

All the IP addresses, everything.

Yes, everything. No, no, you had to actually enter your username. Oh God, passwords. And there's things like, is your pet so wonderful that you use their name as a password?

Yeah.

What's your pet's name? Love it.

Let's flip it.

This is not hard. You don't need like, you know, huge cybersecurity sensors to block this stuff.

Okay, wait, flip it around for a second. There's something going on. I don't know if you read, you probably did. There's an explosion of deepfake porn in South Korea going on right now.

That's not my area of specialization. No, no, no. You guys might know more about that. I heard from a friend. I don't have time for that kind of stuff.

No, no, no, no, I meant you read on Twitter.

And Chamath searched for himself and there wasn't any.

No, but so there's all this fake content that's going to emerge. There's going to be all this clever— Is it fake?

Somebody's fake is somebody's reality.

Yeah, no, but my point is more different. How do you, if you're asked by a customer, tell me if that's real or not, how do you figure out tomorrow if something is real?

Well, look, there's a huge conversation going on that there needs to be some form of regulation that insists that watermarking needs to happen. If you generate a video using any AI tool, it has to say created by AI. If it doesn't, it's very hard to tell the difference.

Yeah.

And as I was saying to somebody else the other day, it's like most likely if it seems too perfect, it was probably created by AI.

How is that different? Let me just ask philosophical. How is that different than airbrushing in Photoshop and making the person look completely different? We don't have any of those disclosures today. I always feel like there's a spectrum that everyone kind of talks about.

I guess scale would be the issue, right? And fidelity?

I don't know.

Sorry?

Scale and fidelity, like, you know, the number of people who can do what you're saying is 0.1% of the population or 1%. Now it's 100.

I think if it's with the intent to deceive.

I see.

Yeah.

Anyway.

And then what about on the other side in terms of defense? Have you started to make AI, you know, shields that—

Yes. Yes. So look, the 2 biggest risks today in AI are, is that I think about 20 to 30% of most companies have employees on the younger side. who are using AI apps to try and get their job done faster and easier. Write me a marketing blog, try and figure something out, write me a script for this, or take this data, analyze it for me. The risk there is that you're sending proprietary data up into a model for a company.

Oh, yeah.

Here's a napkin drawing of a chip design I just made for inferencing. Turn this into a real CAD drawing for me. They could do it, but except that LLM was brought down from Hugging Face and it's going back to North Korea.

Yep.

So there's that risk that your employees are being targeted, the AI apps in your company who are uploading proprietary data in a happy way, very nicely for you. Same guy who picked up the USB stick. So there you have— we have a product that watches all these apps and makes sure that what you're using, what you're uploading, is not being sent to dangerous apps. Or if you don't want your employees to send it, we'll block you. The other one, which is kind of interesting, is that I think almost every company is experimenting with deploying LLMs. internally because they all want their favorite proprietary chat interface. And there you need to be careful because you can— what you used to do with SQL injection, you can do with prompt injection. You can bombard models, you can bias them, you can do a whole bunch of stuff. So we have what we call an AI firewall that'll protect you.

I want to be sensitive to time because I know you have to fly to Geneva. Thank you very much for coming. Ladies and gentlemen, Nikesh Arora.

Thank you for having me.