0:00
Overnight success 3000 double That's misinformation.
Search this video
0:48
>> Clearing order inbound. >> Let's just roll.
0:57
We are surrounded by journalists. Hold your position. Come. Get up. Trust the experts. Five. We are experts. Founder moment. Five good.
1:30
>> I see multiple journalists on the horizon. Stand by. >> UAV online. >> Blaze. >> Double blaze. Triple blades. Double kill. Fight. Cop. team deathmatch. We are experts. Triple blades. Let's just roll. Right.
2:55
Marky clearing order [music] inbound. >> Come get up. Strike one. >> Strike two.
3:30
[music] >> Activate gold retriever mode.
3:46
market clearing order inbound. >> 5.
4:07
>> I see multiple journalists on the horizon. founder. >> You're watching TVN.
4:24
We are live from Falcon Con here with Jordan Curts from Crowd Strike. Welcome to the show. Thank you so much.
4:29
Always great to be back here with you.
4:32
>> Always great to have you.
4:32
Uh congratulations on everything so far.
4:35
I'd love for you to start by taking us through the big announcements today.
4:38
Well, big announcements uh just coming out of my keynote today.
4:40
First, we started with our >> Falcon Guardian product, which is in the category of aid.
4:46
Y So, if we think about the category that CrowdStrike really helped create around endpoint detection response, now this is taking really what we do for a human and a computer and applying it to AI agents. >> Yep.
4:58
>> Obviously, AI agents are more sophisticated.
5:00
They have access to data.
5:01
They have access to compute, to networking resources. They do bad things. >> Yeah.
5:05
>> Cuz they're like a bunch of drunk interns that [laughter] you put on your network. >> Yeah.
5:07
and um and >> they can do a lot of damage.
5:10
>> They can do a lot of damage. Right.
5:10
So, we built this technology.
5:11
We talked to our customers.
5:13
We worked with um some of our largest customers to really figure out what they wanted like Amazon. >> Yeah.
5:20
>> And uh we're super excited because the missing link to accelerating security, sorry, AI adoption is security. >> Yep.
5:28
>> Customers for the first time want to go faster somewhere and they need security to go faster as opposed to a brake pedal.
5:33
It's actually a gas pedal. >> Yeah.
5:35
If you go back throughout your career, I mean, you've been working in security for decades, and I've just been really struggling with the order of operations for the predictions that have come true.
5:45
Like, we got the hugging face attack, and then like 3 weeks later, someone was able to use chat GPT to book a haircut.
5:52
And it's just weird that we're getting capabilities in this very spiky way.
5:56
Does this match your expectations?
5:59
How does what you're seeing today go back to your original thesis when you started the company? Well, it's interesting.
6:05
When I first started the company, it was really based upon we called AI, but it was machine learning back then.
6:10
But again, it was being more predictive and using um the algorithms to figure out whether something was good or bad. >> Yeah.
6:18
>> And now, if you fast forward to generative AI and agents, you know, it's important to to leverage all of the data that we've accumulated over the last 15 years to get to a point where we're still doing that.
6:27
And one of the the second announcements we made was really the super intelligence cyber lab. Yeah. This was very exciting.
6:33
And we actually partnered with Nvidia to create really the first what I would call the agentic security um platform. >> Yep.
6:44
>> That is focused on a red, a blue and a harness that continually learns from each other. Yeah.
6:50
>> So we based um some of our models on Neotron, >> right?
6:54
We can probably get to that, but just overall announcement.
6:55
So if you think about what we're doing, it's really creating a frontier type model and harness at frontier levels specifically built for just the defenders. >> Yes.
7:07
So uh with that uh super cyber cyber super super intelligence lab uh you're going to have more tailored models, more specification probably spikes that go beyond what's available with the stock frontier products I imagine over time uh if not immediately.
7:23
But I'm interested in the uh the benefits of open-source, the benefits of uh thinking economically because I imagine that part of the battle between attackers and defenders is starting to become economic.
7:35
How much how many megawatts can I put behind this attack?
7:39
Is that what's happening?
7:40
>> Well, the compute the the limitation, you're exactly right.
7:42
The limitation for these attacks are really going to come down to sort of compute and cost. >> Yeah.
7:47
>> Because the knowledge, and I talked about this in my keynote, has now been democratized where whether it's a activist, an e-crime actor, a nation state, they're they're all now equivalent because of agentic technology. Yep.
8:00
>> And we called it the rise of the agent state. >> Yes. >> Right. >> Yes.
8:03
>> Meaning that now >> postnation state. >> Postnation state. Right.
8:05
So the apex predator is now >> the the agent state. >> Yes.
8:10
And I think what's really interesting is and it kind of gets back to your spikes and the in these models and those sort of things is that when you go back to hugging face >> it's a fascinating read.
8:18
You probably have have read it and there's some great papers.
8:22
We were actually called in >> to to help open AI go through some of it.
8:26
I'll leave it at that but the papers are out there >> and when you [snorts] look at what these agents were capable of doing >> and just how sophisticated they were with note-taking and communicating and maybe working as part of the collective, it's incredible. >> Yeah.
8:39
But the thing that I really called out in my keynote was that's all groundbreaking, but really what was the aha moment?
8:44
The aha moment was the same incredible frontier caliber AI was available for the adversary.
8:50
In this particular case that were, you know, it was trying to pass a test, >> but it wasn't available drunk interns >> for the defenders. Yeah.
8:57
And and these spikes come with the model refusals and all the guardrailing that that was put in place.
9:03
So again, our models are specifically trained on our data, which we think is a huge advantage. harnesses that we built. Yeah.
9:10
>> And again, we fine-tuned uh in the model creation and and post training. >> Yeah.
9:16
>> Uh in concert with NVIDIA to come out with incredible efficacy at the lowest cost.
9:21
And this is what you were just getting to because customers want the best outcome at the lowest cost.
9:26
>> And I want and I want probably AI review aid at every endpoint, at every transaction, every API call, every single possible moment.
9:33
And that can be very expensive if it's running on a really expensive model. Right. >> Right. Yeah.
9:38
So, so our aid is going to again there two separate announcements but you're going to be able to run that from the platform.
9:43
It's in it's in we announced it today but the the red uh tempest which is the name of that model >> and the blue solano which is the defensive model >> or part of the safe mine system. Okay.
9:55
So safe mine you can think about as the harness right >> but those models and harness will be available in the platform >> to make everything smarter and better and faster. Yeah.
10:02
And then we have a trusted access program for customers in our quilt quilt works program that they would be able to use the models directly. >> Yeah. >> Okay.
10:12
So, uh when you read the agent traces from the hugging face uh attack uh do you see anything in there that feels difficult to detect it?
10:20
Because when I look at those I'm I'm like okay they're crossing a line. This is odd.
10:27
This would flag something in me.
10:27
But we've seen examples like the how many Rs are in the word strawberry where AI just kind of falls flat on its face when you're trying to analyze a particular shape of problem.
10:37
Do you how tractable is this problem?
10:39
How optimistic are you about solving it?
10:42
>> Well, it's tractable because when we when we built our system, it was designed to look at sort of these we call them indicators of attack.
10:48
So if you look at the whole attack chain >> and if you saw the keynote this morning, you know, it's just linking all these together.
10:56
So irrespective of like what happened?
11:00
These are these are kind of unknown attack chains. Sure.
11:01
Part of the issue was it was sort of flooded with so much information that I think it was just sort of buried.
11:07
You know what's going on? Is it real? Is it not real?
11:10
>> And what is going to really take place is the defenders have to use AI to be able to >> connect the dots on those which we do.
11:18
But you also then have to strip out all the noise.
11:21
So you have to separate the signal from the noise >> and that's where these sort of models uh perform well.
11:28
I mean obviously there's an incredible amount of attention on AI.
11:29
What is the role of deterministic threat detection these days?
11:34
Is there is there still is that a useful tool in the tool chest?
11:36
Is there are there still advances being made there?
11:40
Because I imagine that for certain problems throwing a more deterministic uh you know security uh structure around something could actually be beneficial.
11:50
>> Well, you have to and if you look at in the hugging phase incident, this was more of a forensics what happened. Sure. >> Right.
11:55
If you think about deterministic um security, it has to be in line and you have to make a decision.
12:00
You know, you have to be right the first time, right?
12:03
First time final as we call it.
12:05
So um I haven't seen any sort of models, if you will, actually stop a breach in its tracks because >> it's, you know, as it's happening, you're kind of looking at data and those sort of things. It's not in line. >> Sure.
12:16
>> So a lot of what the models are good at is sort of sorting out what happened, you know, sifting through lots of data and those sort of things, finding vulnerabilities.
12:22
But um you have to have a deterministic system which is what we built. >> Yeah.
12:27
>> And I think what people maybe get confused on or maybe aren't quite sure is when they hear things like mythos.
12:34
It's like well you know this super powerful model is going to hack the world.
12:38
>> It really hasn't come up with a new invention of hacking. >> Sure. >> Okay. This is very important.
12:42
It's come up with it can find more vulnerabilities.
12:45
So more of >> and faster. Right.
12:46
And when you combine those, you have and you combine and link these things together, you have greater success.
12:53
But they haven't invented a new way to hack.
12:55
>> Yeah, >> this is very important because I think the the public may look at this and say, "Well, Jesus, a super weapon that can hack anything."
13:02
And it it really is the same techniques.
13:04
It's just more of it and it it can keep track of more things.
13:07
>> In turn is like you're taking a a 10x hacker and making them a thousandx hacker. Correct. Potentially. >> Yeah.
13:11
It's it's almost like Iron Man.
13:13
You put the suit on, >> somebody who's smart is a heck of a lot smarter. Yeah.
13:17
>> So, how how are the threats evolving?
13:17
I imag you you've been talking about agent states, which makes sense as a new concept, but I imagine that you still have nation states that are deploying their own agent states, >> correct? >> On their behalf.
13:31
I'm sure the same thing is happening with various, you know, loose loosely tied hacker groups.
13:34
But how is the threat evolving?
13:37
What are you talking with customers about these sort of new threats?
13:41
Well, it's interesting and it it is the agent state is going to help the nation state, the e-crime and the activists, right? Everybody in between.
13:48
So, obviously that's just an umbrella.
13:49
But I think if you look across those different groups, they all have capabilities.
13:53
They're all leveraging things like these openweight models, right?
13:58
We talked about obliteration really taking the guard rails off an openweight model and you go to hugging face now and you can download these obliterated models and you can run them basically on a on a beefy system, right? It's incredible.
14:08
and you put a question in and you're like there's no way it can answer this and it bills like you want a full malware ransomware kit. Boom. It's done. >> Crazy.
14:19
>> So this is part of the issue. This isn't theoretical. It's here.
14:21
And what customers are asking for is we want we want to give visibility into these agents. Got it.
14:27
We want to stop these sort of threats, but we also want to know if it's an AI attack.
14:31
It's actually a very important question that they need answered.
14:36
Is it an AI attack or is it sort of an adversary with maybe AI assisted?
14:40
That's been one of the number one questions because it informs them on what they need to defend against, but it also informs them on how they express this to the rest of the company, the CEO and the board. >> Yeah.
14:50
What's been the biggest or what do you see as the the biggest bottleneck for you in terms of actually deploying solutions to customers?
14:56
They're asking you, do you need more sales reps?
14:58
Do you need more uh just more customers to come to you just scale up what you already have?
15:05
What's the shape of the next 12 months for you?
15:07
>> Uh the the great part of what we built at CrowdStrike is it's a very scalable model.
15:11
It's a single agent, >> single platform with a single control plane.
15:15
>> So, you know what people need to do to roll out aid? >> Turn it on.
15:19
>> You know, sign the [laughter] PO. I need to turn it on. That's it. Right.
15:21
It's the same agent that's there.
15:24
>> And we spend a lot of time >> um making sure that we can instrument, we can find shadow AI. Is it claw? Is it CEX?
15:31
whatever it is, >> we can instrument every action that agent has taken, every action >> with a specific identity, every tool call, every spawn of an agent, >> every network connection, every prompt. We have that visibility. It's incredible.
15:44
>> So, it's sort of like the EDR moment when we developed EDR, >> when we first showed people, they were like, I've never seen this before.
15:50
And when you do this to an agent, they're like, we've never seen this.
15:54
We've been dying to see this.
15:55
So I think when you look at our model, we're combining a very scalable platform. Turn it on, same agent.
16:02
That's a huge win for us and a huge barrier to entry for other competitors.
16:05
We have the most agent security agents deployed of any pure play security company. So that's one.
16:11
And two, we combine that with a very flexible licensing model. Yep. >> Which is Falcon Flex.
16:15
So you'll be able to to use it's a it's a tokenbased system.
16:19
So the more AI you use, the more you pay >> um because there's, you know, more cost to this.
16:24
But at the end of the day, you'll be able to use those credits and burn down from your Falcon Flex licensing model. >> Yeah.
16:30
And I imagine with Neotron and just the advances in models like there's a world where token prices come down over time if if you know you're scaling up and stuff.
16:38
So there's a lot of flexibility there.
16:40
>> Well, and we we have different models which make it a lot more cost efficient.
16:44
So the the goal for us is to really Yeah.
16:45
is actually to really it is it's not just one model [clears throat] >> of course.
16:48
Um, so we're we're really driving down the cost and I think that's a huge >> uh advantage for our customers and then providing that sort of trusted access.
16:56
We've been doing this a long time and customers want to retain their data and the the provenence of that data, the sovereignty of that data with us. >> Yeah.
17:04
Take me uh a couple quarters for it, a couple years for it.
17:08
As far as you can because it feels like we there was there were sci-fi stories about cyber security incidents related to AI.
17:15
Then we got the mythos moment, the hugging face moment.
17:19
Now we have a really solid response and it feels like there's, you know, the the never- ending cold war continues.
17:26
But are are we are we in a stable equilibrium here?
17:28
Are you expecting some big change one way or another in the posture between the two waring groups here, the red team, the blue team?
17:37
>> I think it goes back to >> the story as old as time. It's good versus evil.
17:41
[laughter] >> It really is.
17:43
This just plays out now in the modern day with agents at a speed that we can never really contemplate. >> But we can fight.
17:50
>> But we can fight and we will. They'll get better. We get better.
17:53
>> And you know, part of part of our what we delivered today with the lab is the red blue training loop.
17:57
Very important >> is that the blue learns from the red, right?
18:01
So the defensive model continually learns from the offensive model and you have a very fast cycle. It's very important.
18:07
But >> you know there's going to be all kinds of new technologies, new agents, new systems, things that we haven't even heard of today. Yeah.
18:13
>> And we have to be able to defend against that.
18:16
And I think what remains while there's will be a lot of change, what remains constant is security >> parallels the slope of the technology curve. >> Yeah.
18:25
>> So the technology curve, I mean, >> you know, I started in the early 90s doing this, right?
18:28
It was like this and then it it's like that.
18:30
So you have to have security that actually parallels that.
18:34
>> Um, we don't do everything.
18:34
You know, what we do, we do really well.
18:36
We're a big platform company.
18:38
There's only a few of us.
18:39
I think that's going to win in this market.
18:41
Um, but it's a big market.
18:43
You can see by all the the companies around here. >> Yeah.
18:45
And this is like a for those of you that are tuning in, this is a uh it feels like you set up a town here, like 10,000 people. It's massive. >> It's unbelievable. Yeah.
18:53
I mean, you know, maybe you'll see some of this in B-roll or whatever, but >> this is a massive security conference.
19:00
We have companies coming to this going, "We're not going to any other conference."
19:03
And it was just an offshoot because we've got the best customers. It's a big audience.
19:06
But I think what's important to realize is we understand and value the ecosystem.
19:10
We can't do everything like what we do we do really well but it's part of the whole ecosystem and network which is why >> uh Nvidia is here obviously Jensen Wong this morning lian from Intel Greg Brockman from open AI all partners plus all the many that you see here. >> Yeah. Uh sorry Jordies please.
19:26
what uh what groups or institutions are not paying enough attention to this new technology cycle that uh everyone here is like paying attention to this obviously the AI boom new threats things like that but is is there a set of groups globally that that need to be u paying more attention to the new set of threats that aren't today?
19:49
>> Well, it's a good question and I think the mythos moment has really provided uh much more visibility from the board all the way down to the CEO level.
19:55
I mean, my phone was ringing off the hook from Fortune 10 CEOs going, "Hey, what does this mean?
20:00
How can you help us, etc. , right?"
20:01
So, you know, from visibility standpoint, that's good.
20:03
And then you look at a Fortune, we'll call it a Fortune 500 company.
20:08
For the most part, they have or will find the money to deal with some of this and depends on the industry, how much they spend, but generally they have a view and it's regulated, etc.
20:18
The have nots, those are the half.
20:21
The have nots are >> I'm thinking like local utilities, hospital systems, >> hospitals, NOS's like utilities, forget it.
20:30
I mean, they're running such old software.
20:32
>> So, it's the have and the have nots.
20:32
And I think part of what we want to do and even working with Open AI is how do we help, you know, give a hand up to people >> who, you know, need it because they don't have all of the security people they need.
20:45
They don't have the money for all of these sort of advanced software and technologies.
20:49
But we got to it's a collective community effort and that's part of where what we're helping to drive in partnership with many others.
20:56
>> What does it take to make it a crowd strike these days?
20:57
You're hiring uh AI researchers now for the new uh super intelligence lab.
21:01
You have a lab >> with 270 PhDs. >> Wow. Uh significant.
21:04
So uh what is the shape of the new allstar upandcomer at CrowdStrike look like?
21:14
um the upandcomer I mean it depends on the >> I guess the question is just like how much AI are they using how much are the human skills still hyper relevant what is the what is the balance how how familiar do you have to be with these tools what are the pitfalls because I think everyone's sort of realizing as they run large companies that people can get lost in the sauce if they're using
21:33
too reliant on it how do you think about this in terms of like your own management style >> yeah I think that's important because I mean we we try to be very deliberate about it security is very important to Uh so where we use it, how we use it, um what groups we use it in, you know, we've expanded out obviously, but >> obviously a big part is going to be around coding. >> Sure. >> Sure.
21:52
>> And you have to make sure that you get the right secure code out of it. >> Yeah.
21:57
>> You [clears throat] know, it used to be in the early models, I would say a little less so now, but in the early models, it was like the early days when coding when someone would go out to the internet and they would just copy and paste a snippet of code.
22:07
paste a snippet of code. you would take that vulnerability and would propagate for everybody that needed you know that snippet of code right so AI was originally generating some code and you're you know I I have my own models I built I you know have my own security agents that I built just to play around >> and it's like well the same model that just built my code then I built a an
22:28
agent to figure out whether it was secure was the same model that goes it's not secure like okay why don't you build it you know in the first place so you have to be aware of that and but what I think is important getting back to your question is if you don't buy into AI as an enabling technology and you're sort of scared for your job, you're not going to you're not going to be successful at CrowdStrike. If you want to use AI in
22:47
If you want to use AI in the right places at the right time with the right cost, >> we have all the room in the world for you here. >> I love it.
22:55
>> And we have these sort of >> AI builders and we're deploying them into all the different functions.
23:00
So those are the folks that you go, "Hey, I it'd be great if we can do this."
23:04
And and you turn around and get some coffee and they go, "Here, it's done."
23:07
Yeah, that's what we like. No, I love that too. I love that, too.
23:10
Even in our small organization, we would love for you to sign this helmet. We have a Sharpie here.
23:15
Uh, would you mind signing right here?
23:17
We want to get an autograph from you to commemorate the occasion.
23:18
And we also have a gong with a mallet.
23:21
We'd love to >> get you to smash this for >> the occasion. Here you go. Give us a gong head.
23:29
We'll let you get a backhand on this, right?
23:32
Where's the the sweet spot?
23:35
>> I think the sweet spot's right here.
23:35
No, just give it enough force. You'll be good. >> All right. >> With authority. >> There we go. With authority. >> Well, thank you.
23:44
>> And I just want to say I love the I love the aesthetics of everything here.
23:46
We got Agents of Chaos back here. It's incredible. >> It's incredible.
23:50
Well, congratulations to you cuz I know >> you you had a new baby. >> I did. That's breaking news. [laughter] >> Okay. >> I had a new baby. No, no, it's great. It's not a secret. >> Did I scoop you? >> You scooped me. [laughter] Okay. Sorry about that. >> No, no, no. It's great.
24:02
Great to see you, George. on the track.
24:07
>> I'm going to leave it.
24:07
I'm going to leave it, >> but I Yeah, I won't walk away like I normally do.
24:10
Um, we'll see you at the track. >> Yeah.
24:13
>> Uh, stay tuned for the race this weekend. >> We're excited.
24:15
>> Hopefully our men our men do well and uh we'll go from there. Okay. All right. Yeah. Bye. See you soon. We'll talk to you soon. >> Cheers.
24:22
>> Uh, that was fantastic.
24:22
Of course, the show is >> breaking news >> sponsored by RAMP. Time is money. Save both.
24:28
Easy use corporate cards, bill pay, accounting, and a whole lot more all in one place.
24:31
Thank you, Ram, for making TBPN possible.
24:33
We have a bunch more guests coming on from Falcon Con here in Las Vegas.
24:37
And I believe we have our next guest ready to rock.
24:39
We have Michael Santonus, the president of Crowd Strike, here to take us deeper on everything Crowd Strike is doing in cyber security in the age of AI, the AI revolution is here. Are you ready?
24:51
That's what it says outside.
24:51
Little ominous, but probably accurate. >> Accurate.
24:55
>> Good question to be asking. How you doing, John? Pleasure. Nice to meet you. >> Welcome to the show.
25:00
>> Uh we're going to have you throw this headset on.
25:02
Uh we'll get this out of the way, get comfortable, and uh we will ask that the microphone just go a little bit close to your mouth because it's noisy in here.
25:12
You've got a lot of lot of partners here.
25:14
It's a huge company, huge conference.
25:16
Um did you have anything to do with that?
25:18
Are you the reason why everybody's here? >> Yeah, a little bit.
25:22
>> A little bit of a reason. Okay. Yeah. Yeah.
25:24
Well, yeah, let's start with your role.
25:26
Uh what do you do at CrowdStrike?
25:29
Little bit of your history, and then there's a whole bunch of hot topics I'd love to go into.
25:32
Happy to go wherever you want.
25:34
>> Let's start with uh your role, your day-to-day.
25:37
>> President of Crowd Strike.
25:37
So uh go to market reports into me.
25:39
Product and engineering team.
25:41
Uh researchers, our threat hunters, the >> researchers too. >> Researchers too.
25:47
>> Researchers and go to market.
25:47
How are those feel like different groups?
25:49
What how how does that work? Is that uh operational?
25:52
Are you spending more time with one or the other? How does that blend?
25:55
Um, look, I think uh my background I I started through the the product side of the the organization.
26:04
>> Uh, I was the CTO of the company for for quite a while.
26:06
One of the things that I talk a lot about is you can build the best technology in the industry.
26:11
>> If you can't sell it, if you can't talk about the value, if people can't deploy it and they can't ultimately keep themselves safe, >> it's kind of irrelevant.
26:18
So, it's trying to bring together >> uh the smartest people.
26:22
Obviously, you had George on before.
26:23
We worked really closely.
26:25
He's got a bunch of people that report in to him as well on the engineering and research side.
26:30
>> It's bringing it all together and then [music] making sure that we keep people safe and secure. >> Yeah.
26:33
What are what are customers in the go to market side actually concerned about because there's an immense amount of attention on cyber security right now.
26:41
I imagine that makes sales easier but at the same time we've all seen token maxing and AI budgets and people you know they have to find the money somewhere.
26:49
So what is the tension that you deal with?
26:51
How are you thinking about positioning that for customers?
26:54
Yeah, look, every organization has huge uh requirements, but they've also got a budget.
26:58
They've got a business to run.
27:01
The business is not there unless you're a service provider and your business is cyber.
27:04
You're building cars, you're building houses, you're in medicine.
27:09
>> Um, and you need to keep yourself safe and secure.
27:11
So, every CISO, every CIO is trying to basically do the best at what they can.
27:16
they they came into 2026, suddenly everyone's trying to work out how do I find tokens to pay for all of the things that the business is doing.
27:25
>> You know, every second organization has token chalk.
27:27
So, at the end of every month, they just realize that they spent 20,000 more than they should have per employee.
27:34
>> Uh, and you know, we got to work with everybody to show them there's a better way.
27:38
There's a more efficient way.
27:38
We give them a a a vehicle that they can procure and get Crowd Strike Flex that I think you guys have talked about with George in the past. Yeah.
27:47
>> And we show them [music] that they get a much better solution.
27:50
It's just easier to live with dayto-day.
27:51
So it's not only showing them the cost of buying it, but how do they deploy it and then operationalize it? >> Yeah.
27:59
>> Uh because cyber you got to live with it.
28:01
You got to use it every day. >> Yeah. Yeah. Okay.
28:02
So in deployment, uh let's talk about the balance between attackers and defenders.
28:06
In terms of AI capabilities, I'm pretty confident that you, Open AI, Enthropic, the big labs have the advantage on raw power, intelligence, capabilities in terms of defense and probably offense, too, but you don't use it for that, [laughter] you know.
28:22
Um, but does that map to your reality that that the >> only the only difference is like you guys need to be successful every single time and an attacker only needs to be successful one out of it. >> Yeah. Yeah. >> 100,000 times, right? >> Yeah.
28:35
>> I mean, that's the way it works. >> Okay.
28:37
Um, every organization can't get it wrong and equally they got additional pressure.
28:41
Yeah, they got to get it right 100% of the time, but they also have to not stop business.
28:45
Yeah, >> it's easy to get it right 100% of the time if you turn off everything.
28:50
>> So, you know, it's not going to it's not going to fly with the CEO of the company and if every day, you know, the security technology is slowing people from browsing, you can't use the AI that you want.
29:00
So, there's an added level of complexity.
29:04
>> The defender has things like change control.
29:06
The defender has things like you know regulatory uh guidelines and process that they need to you know follow.
29:13
Attacker doesn't care about any of that.
29:14
You know you go to Europe where everyone in Europe is talking about like AI laws and privacy and regulation.
29:18
I said at a conference this is fantastic.
29:23
You guys are leading the world in all of this regulation and everyone was proud.
29:26
The adversary does not care. [laughter] >> Yeah. >> The adversary. >> Yeah.
29:30
They're already breaking the law.
29:31
So why would they not why would they follow >> regulations? They love it.
29:33
they're in there and and that's that's kind of some of the challenges that people face with and then you have the cost pressures. >> Yeah.
29:40
>> You can't just pour everything into tokens.
29:43
>> You can't just pour everything into, you know, it budget.
29:46
>> So that's why we spend a lot of time making sure that you get the best that you can.
29:50
>> And a lot of the time we come in and say, "Okay, here's a product.
29:51
We're going to try to take two out." >> Yeah.
29:54
>> It's not one in, one out.
29:54
It's one in, two out, three out. >> Sure. Sure. Sure.
29:57
Um but so uh talk about the pace of AI diffusion AI adoption in uh in cyber because it feels like even though there's immense go to market operations AI adoption among the bad guys has to be incredibly quick because it's often free.
30:16
It's it's you know highly motivated.
30:19
There's no just download the thing.
30:21
Um and so I feel like that's maybe more of the challenge than like raw intelligence.
30:25
Is that a reasonable frame of mind to think about the the real showcape of the problem is like the slope of adoption like the the good guys need to adopt more security AI faster than the than the than the attackers.
30:40
>> Look, we've talked about this for 20 plus years.
30:42
If you went to a security conference 20 years ago, everybody talked about this time in the future is going to come. >> Sure.
30:48
>> Where attackers are finding vulnerabilities and they're weaponizing them at a speed that you're just not going to be able to deal with. >> Sure. We're here now. >> Yeah.
30:55
[laughter] And if you think about the technology, I mean the adversaries get access to everything.
30:59
And the best thing that's happened to them is the advancement in the open models.
31:02
They can go and get a Chinese model.
31:05
They can go and get a model effectively from anywhere that they run inside their their sort of framework.
31:11
Uh they can build it, they can train it, they can tune it.
31:15
And because they're running it in their environment, they can customize it in a way that you don't know what they're doing. >> Yeah.
31:21
>> And then the first time they use it is the first time you have to deal with it.
31:25
And it's cheap for them when they run it internally.
31:27
They don't have the cost of the tokens and everything else. They need the hardware.
31:31
They need everything out there. >> Yeah.
31:32
Do you do you feel like their adversaries are computed right now or are they just able to do a lot with a little?
31:39
>> Depends on the adversary.
31:39
You know, the kid at home is going to be compute constrained.
31:43
uh if you're dealing with a nation state, if you're if you're dealing with some of the most uh wellunded, well structured, well-trained adversaries that now have this capability as well, this is the challenge.
31:54
That's why there was a packed uh arena this morning.
31:56
That's why Jensen came out to talk about what they can do because if we don't do something and if it's not a community, it's going to be really hard to compete with this technology. It's so good. >> Yeah. Yeah.
32:07
What uh what kind of conversation are you having with governments these days?
32:10
Uh we were just talking about uh you know it's not just enough to secure the Fortune 10, the Fortune 500.
32:17
You need to secure the local hospital, the local uh you know utilities operation.
32:22
And it feels like the government, various governments might try and incentivize different rollouts and speed things up there.
32:29
What are you hearing from talking to people in government?
32:33
Generally, >> I think I think you've nailed the point because the world's business runs a small business.
32:39
small business. the world's economy I should say runs a small business they don't have the resource they don't have dedicated people they don't have the the dollars and most often when they have a problem it's also very hard where do they go and they think you know I'm not going to be able to get the resource of
32:55
the largest banks etc u we want to spend a lot of time with them and that that's super important but importantly governments around the world that are also starting to say hey we need to make sure that our economy keeps working we need to make sure that critical infrastructure you know the power is on, water is flowing, um you know, rubbish trucks arrive. Um otherwise, we're going
33:12
Um otherwise, we're going to have have chaos.
33:14
And you know, we've seen examples of critical infrastructure taken out.
33:20
We can't have banks get compromised with ransomware.
33:22
We can't have uh you know, just pure uh chaos out there.
33:27
And I think there's an opportunity now to work more collaboratively >> and every every coming coming back to your question, every government wants to talk about AI, you know, what do we have to worry about? How do we use it? How do we embrace it?
33:38
But what do we have to be worried about when other people that have malicious intent? >> Yeah.
33:43
>> Uh what could they do to us?
33:43
You know, if they start using all of these openweight models, uh what what do we need to know about them?
33:49
So, a lot of partnership, a lot of collaboration.
33:51
Um and I think, you know, that concept of community that we're talking a lot about this week is super important because if we don't do that, we're just not going to be on top of it.
34:02
>> Tons of attention on AI for very good reasons.
34:04
uh is quantum getting under discussed and I say that because I've heard that there are adversaries who are hoovering up encrypted data with the hope that quantum computers in the future will be able to decrypt that data. They're stealing it now.
34:23
They can't do anything with it.
34:25
And so maybe even though AI is super important, we should be talking about it 99% of the time.
34:30
Should we be talking about quantum 1% of the time?
34:32
talking about quantum 1% of the time? uh we are and and it's look it's a big topic um not a day go by where a customer a partner an analyst someone asks a question about quantum and what that means >> I love the point um just to kind of dive in where people are taking data
34:48
>> uh a lot of the time you see uh examples where attackers will basically excfiltrate all your data >> and then you don't see it again >> and you get the email saying uh yes your name was in clear text but everything else was encrypted and now you have to think in 5 years that might be decrypted Sure. >> But but sometimes when those attacks
35:03
>> But but sometimes when those attacks happen, you understand when when you see your name in the list, you understand what they were doing.
35:08
You understand that you are the person that they're monetizing. >> Sure. Sure. Sure.
35:12
>> But what happens when they take terabytes of data, they're not selling it, they're not acting on it, they're doing something, >> they don't even know what it is, maybe.
35:18
>> Well, they didn't do it just because they wanted to have fun. >> So there's intent.
35:21
Now, whether it's decrypting it down the road, whether it's training models, you know, you need to think every attacker, there's a motivation for it.
35:29
And and it's fascinating when you actually start to get behind what's interesting what's what's driving them. >> Yeah. Yeah. Yeah.
35:36
>> You got to imagine there's at least a few people out there that [laughter] >> Well, that's how that's how it used to be like back in the day.
35:42
You wrote malware because you wanted a company like Crowd Strike to say, "Hey man, that Jordy guy, he he >> Oh, yeah.
35:49
The cloud >> like he nailed it. Like this is an attack.
35:51
This is really innovative. We talked about it.
35:53
We started you you felt good about it.
35:56
You got it in the hard place.
35:57
>> Doesn't work that way anymore. >> Okay. Yeah. Two two.
35:58
How uh how is AI impacting various social engineering schemes?
36:05
>> Well, it's funny because you know a few years ago it was really easy to tell people about social engineering and said you know if you read something and it read you know reads like you know a 10-year-old wrote it or someone with bad English you know it's probably not the bank that you have all your money in right it was an easy telltale sign. Yeah.
36:25
>> Uh now they they learn how to use uh >> you know chat GBT, they learn how to use Gemini.
36:29
The the emails that they write are phenomenal.
36:32
We used to see about a year ago the click-through rate in fishing was about 11 to 12%.
36:38
>> Today with AI that still feels incredibly high.
36:40
It's over 60 now [laughter] >> because written perfect grammatically they probably write better than us now.
36:49
>> And you kind of look at what comes out.
36:50
It's really hard to work out what's real, what's not.
36:52
So they're getting a lot of opportunity.
36:54
The thing is you grab an openweight model and you you basically say how how would I carry out this attack.
37:01
>> It's going to give you the playbook.
37:03
>> You know, I've I've done cyber since university.
37:05
You don't need any of that anymore.
37:07
I kind of feel like wasted youth now because you just need a model.
37:10
You ask the question >> and it tells you what to do. >> Yeah.
37:13
Uh the company's on Okay, we got to move to the next.
37:16
We got one more question.
37:17
We got time for one more question.
37:18
I mean the company's on an absolute tear.
37:19
I'm interested in how you know you oversee all these groups.
37:22
What were you telling people during the SAS apocalypse in earlier moments even go back further just just throughout the company's history?
37:30
You've had a very clear vision of where things are going, the value that you're creating long term, but there's girrations.
37:35
What is it like actually managing all of these different teams?
37:39
different teams? It's got it's got to be funny that like the the height of the SAS apocalypse, you probably had the phone ringing off the hook more than any other point in history because at the same time people were like realizing
37:50
like wa models are now at the point where >> um >> look I I've known George who he had on just before for for over 20 years and I remember one of the first things he ever said to me look after the customer everything else takes care of itself. >> And that mantra goes throughout Crowd
38:03
>> And that mantra goes throughout Crowd Strike.
38:06
So all of this noise, you know, we just basically say keep looking after the customer, keep innovating, keeping them safe and secure, things will take care of themselves.
38:14
>> We know the way the technology works.
38:17
You know, for me, the SAS apocalypse thing, there was no merit to it.
38:19
It didn't make any sense at all.
38:21
Even today, when people say, "Hey, all the models are going to find all the vulnerabilities and then we're going to get this state of normality." No, we're not.
38:28
We're going to get more vulnerabilities.
38:30
We're going to get more attacks.
38:31
It's only going to get harder.
38:33
And that comes from just having so many years of experience.
38:35
It's having an engineering team that is at the cutting edge at the forefront.
38:41
>> But it's putting the customer first and you know it's a it's a good formula and it always works. >> I love it.
38:45
Well, thank you so much for the show. Have a great day.
38:49
Congratulations on fantastic Falcon.
38:49
Up next we have Daniel Bernard, the chief business officer of 5 minutes though.
38:57
>> Five minutes to hang out.
38:58
>> 60% clickthrough rate on fishing emails. What are you guys doing? What are you doing?
39:02
[laughter] What are you guys doing?
39:04
Uh, speaking of fishing emails, there have been a raft of uh of uh password reset attempts on X.
39:12
>> We got another minute I wanted to ask him about. >> Oh, yeah. Yeah.
39:14
Well, we can we can go into that with the next guest.
39:16
Uh, Nick Carter posted on X.
39:17
A lot of people getting unsolicited X password reset attempts in their email inbox. Do the following.
39:23
Go to X settings, security, account access, security, check password reset protection.
39:28
Don't let people hack into your X account.
39:31
It's too simply too valuable.
39:33
You can't let >> And this was because uh X money is rolled out to a lot more people notification financial >> incentive now because you could potentially steal someone's money if you got in as opposed to just post a Beamcoin link or something like that.
39:45
Uh anyway, the other story we got to talk about is uh uh YouTube creator director Markiplier has acquired an 8 8. 5% stake in GoPro. Did you see this? The action camera maker.
39:58
They've been sort of uh in the doldrums.
40:01
A lot of competition from China, but uh >> and then immediate questions because GoPro just got acquired today. >> Oh, wait.
40:07
It was a full acquisition.
40:09
>> Yeah, it's being >> Oh, I didn't know that. >> Wow. >> Uh yeah.
40:12
So, this news comes out yesterday and uh just today GoPro has entered into a definitive agreement to merge with privately held Starman Optical in a deal valued at 285 million.
40:27
So it seems like Markiplier is up uh massively which is going to immediately >> like part of the deal. Yeah.
40:34
>> Well, it's going to immediately draw a lot of attention.
40:36
Um >> yeah, >> but a lot of times these things happen as like one piece of a larger deal and the state gets disclosed at a certain time because it's part of this remaking of the business.
40:45
Uh do you think go can come back?
40:50
>> That's a good question.
40:50
I >> I I owned a GoPro back in the day.
40:52
I never I never GoPro was one of those things where for the average person, you're capturing footage that is only entertaining to you and no one else is going to care. >> Yeah.
41:04
>> Um, you know, I'm a I'm a pretty good snowboarder, pretty good surfer, >> which is pretty good.
41:09
>> Um, [laughter] remember Shawn Magcguire Shawn Magcguire was like [laughter] throwing throwing shots, but uh we'll we'll set up a heat, Sean.
41:17
Um but um uh but but but anyways, I I always felt like I would I would film something with my GoPro and then uh it would be mildly entertaining for me and not that entertaining for someone else.
41:31
>> So you don't watch game footage.
41:31
You don't get out there on the waves and rewatch.
41:34
>> If you're watching If you're watching game footage, it's better to watch from the third person.
41:38
>> You don't want to watch the first person. Interesting. Interesting. Yeah.
41:42
>> I I I've owned a GoPro at at various points in time, but again, like never really found a good use for it. >> Yeah. Yeah.
41:47
And it it felt like the GoPro budget for consumers shifted to drones.
41:51
They didn't they didn't get there.
41:53
Again, that's a third person view that is like I think a lot more interesting to >> and there's even some drones that will follow you out while you're surfing and track you well and stuff. Yeah.
42:00
Uh and then yeah, just the innovation >> like I I still have very positive feelings towards GoPro.
42:06
They they work with so many amazing athletes over the years. They were a pioneer.
42:10
Uh the big the bigger challenge was just the iPhone.
42:15
the guy got very durable and the and the quality got amazing.
42:17
So, why would I can take you can take your phone out on a on a ski run or whatever.
42:24
>> Um and uh and you'll be fine.
42:24
The other thing like meta glasses too meta uh the Oakley Meta glasses.
42:31
>> I think Best Buy reported earnings and said that smart glasses are like driving significant instore sales for them.
42:35
Like they're they're actually moving. They're selling well.
42:38
Uh we haven't we've seen that. >> Yeah.
42:40
So So anyways, I don't like to see Insta 360. >> Yeah. and >> DJI take over. So, I hope they can.
42:48
>> But if you watch the independent product reviewers, like those products have innovated in a many in many many ways that uh that GoPro has not been able to keep up with mostly because of the manufacturing uh side of the business.
42:59
But, uh interesting to see, you know, is this Markiplier's way of like buying a new merch line?
43:04
That's one potential view on this is like, okay, you buy a stake in this and then you run you basically run ads on your platforms to promote the new products.
43:13
Uh maybe he has a vision.
43:15
I mean, he didn't make a whole movie uh with a lot of VFX.
43:16
Maybe he wants to uh grow GoPro into something that's more for filmmakers and uh and cinematic >> cinematic creating.
43:28
The uh >> Starman Optical, the acquirer, is an American company that describes itself as a privately held US optical photonix company focused on developing and domestically manufacturing optical transceivers.
43:39
So, when I first saw the news, I assumed it was a a Chinese company uh buying it up, but um but we'll see.
43:48
Hopefully, they can make and sell a lot of GoPros in America. Markiplier.
43:52
I I love his journey and I think uh it you know this feels like an outside of the box move.
43:57
It's not just another uh you know sparkling water brand or hard selzer brand from a from an influencer.
44:04
It's him thinking about business in a very different way. So it's exciting. >> Uh more news.
44:08
>> Uh first let me tell you about Railway.
44:10
Railway is the all-in-one intelligent cloud provider.
44:11
Use your favorite agent to deploy laptop servers, databases and more.
44:14
While Railway automatically takes care of scaling, monitoring and security. >> More news. Mr.
44:18
Beast has launched a book for his audience of voracious readers.
44:26
Uh the people are clamoring.
44:28
>> Uh people in his audience have been asking for a book. He delivered.
44:32
>> Um no, he's turned the book into effectively a lottery.
44:35
He's going to give away a million dollars to somebody that buys the book.
44:38
>> Was that on day one or is that downstream? I guess that's day one.
44:42
That's the that's the promotion.
44:42
So, uh anyway, he partnered with James Patterson who is a huge author.
44:46
Uh, good news for the subset of Mr.
44:50
Beast Beast fans who love to read mystery novels, the YouTube giants collaboration with James Patterson is out Tuesday. Uh, that's today.
44:57
Accompanied by a flurry of marketing on Mr. Beast channels.
44:59
Read my book and you could win $1 million.
45:01
Uh, bad news for Harper Collins.
45:04
The book buying subset of Mr.
45:05
Beast fans appears to be vanishingly small.
45:07
Two people close to the project say The Most Dangerous Games is on track to be a historic bomb with pre-orders numbering in the four digits as of last week.
45:15
But why would there be pre-orders if it hasn't launched the marketing yet?
45:20
This I I'm I'm I'm sort of skeptical.
45:25
>> This is the first time I'm hearing about it.
45:26
>> This is the first time I'm hearing about it and it seems like Mr.
45:27
Beast just uploaded the the actual contest and like you could say she shouldn't be running a lottery or I I I don't like this type of book promote promotion, but like that's I will say if the pre if the pre-orders or the orders still stay in that in that four digits, there's some hedge funds getting involved.
45:46
>> Hedge [laughter] funs buying up buying up more of the >> book to win the million dollars. Yeah. Yeah.
45:50
>> You could win hedge funds.
45:50
You could >> typically with it with a raffle, you don't have to actually buy the book.
45:54
You usually can just sign in and and and send it over.
45:57
But uh will you be reading it?
46:00
>> I think one of us has to read it. Or at least Tyler. >> Tyler can read it.
46:03
>> Where [laughter] are you?
46:04
>> We can get Tyler to read it.
46:04
Anyway, uh we'll have more fun with that in just a minute.
46:07
Let me tell you about public. com.
46:08
Investing for those that take it seriously.
46:09
They got stocks, stocks, options, bonds, crypto, treasuries, and more with great customer service.
46:13
And our next guests are here.
46:16
Welcome to the stage at Falcon Con with TVPN. Great to see you. How are you doing? Great to see you. How you doing?
46:24
>> Well, yes, John, Jordy, uh we're going to have you throw on these headsets and it's a little bit loud in here.
46:29
There's a lot of Crowd Strike fans.
46:33
>> The Crowd Strike fan zone's going insane right now.
46:35
And so, uh just get the microphone sort of as close as you can. Let's move this around. Flip this up. Yeah, there we go. Is it?
46:43
Oh, I think you put it on backwards. >> There we go.
46:46
Anyway, uh let's start with uh introductions. Uh introduce yourself.
46:51
Tell us who you are, what you do.
46:53
>> Hey guys, Daniel Bernard. You can call me DB.
46:55
Chief Chief Business Officer at Crowd Strike.
46:58
>> How many How popular are internal nicknames at Crowd Strike? >> How?
47:01
>> They're super popular, but there's only one DB. >> Okay, that's you. And you? >> Uh Justin Banano.
47:04
Uh I lead the enterprise business at Nvidia.
47:06
So, good to see you guys.
47:08
>> Yes, good to see you again.
47:08
And tell us about the partnership.
47:10
Tell us about the news today. >> Well, big news today.
47:12
We launched Safe Mind Cyber Security's first frontier models and harnesses custom for cyber >> made by Cyber for Cyber.
47:20
We built this on Neotron >> and it's bending the curve of Frontier AI and the and the and the advantage of defenders better harnesses plural.
47:31
>> There's multiple harnesses.
47:32
>> Why would you pick one?
47:32
What what's involved in in selection?
47:34
What are the differences?
47:37
differences? is are we talking about pure economics to tokconomics or are there more like you know right tool for the job >> right go ahead >> yeah let me give a little bit of color I think the the big news too is that the frontier is in the harness >> okay >> um it's not really about just the model it's about the entire system and so what
47:55
the crowd strike team have done a phenomenal job doing is >> tuning the harness for attack tuning the harness for defense um and you know the harness is the thing that's going to sit there and reason and call tools tools and you know work through solving the problem whether it's uh finding vulnerabilities or finding and writing uh detections. Um and so uh I think the
48:14
Um and so uh I think the work that we've done both through the harness and through the open model creates this like super capable agentic system uh that is going to always be on and be able to uh say uh learn from enterprise environments.
48:28
Jensen talked a lot about how we're deploying it internally.
48:31
We're building the digital twin of our environment.
48:33
uh we can uh basically go through these attack defense simulations uh to build uh the best defenses for our organization. >> That's great.
48:41
Um as you went about building this, how important were benchmarks to you?
48:44
Public benchmarks, private benchmarks, like how do you see that fitting into the tool chest of building a great product?
48:52
>> Super important because we needed to be more performant than what's out there today. Yeah.
48:55
Like the goal here that we both set out to achieve is this thing needs to be better, faster, and more cost-effective than the other open source and frontier models of the day for cyber security use cases.
49:06
You know, that's the big thing.
49:07
Like we're not here to change the world of science, math, uh manufacturing.
49:13
We're here to stop breaches.
49:15
We're here to make cyber security better.
49:17
That's offense, that's defense, and that's continuous learning. >> Yep.
49:21
>> Better performance at each step of the way.
49:22
And that's what the data we have that we're able to that we shared with the market today.
49:26
>> Talk about the decision to go with Neimatron.
49:27
There's a lot of open- source models.
49:29
I can imagine why you didn't pick some of them, but break down the decision.
49:32
>> Look, there's a really really close relationship between Crowd Strike and Nvidia.
49:36
So that we didn't even look at anybody else because, [laughter] you know, when it comes to like the foundational layer of AI that we've built the business on, >> you know, that's that's GPUs. Yeah.
49:46
>> And who do we get our GPUs from? The creators of them. It's Nvidia.
49:48
And so when it comes to open source and you can look at NVIDIA and Jensen such a strong perspective and really across >> everything you're doing too Justin like the world needs open source the world needs choice that just aligns with us very culturally as well. >> Yeah.
50:02
>> So >> there was nowhere like why would we go anywhere else >> and and for us the the feeling was mutual.
50:07
Uh I think Jensen said on stage uh CrowdStrike is our number one partner in cyber security.
50:11
They have the perception system that really understands what's going on in customer environments. Yeah. Yeah.
50:16
environments. Yeah. Yeah. So if you pair that perception system with we'll say an open model that we build it's built as general knowledge uh but we we put out there the data sets uh the open techniques and the weights so they can
50:29
be customized uh by crowd strike so they can build their own specific cyber domain intelligence and be able to uh build a new business model where they're like selling tokens right to secure uh enterprises and they're doing it in the most cost effective way by building on that open foundation. What else is
50:44
What else is Nvidia bringing to the table around a project like this?
50:49
Because uh Neimatron is obviously the the model layer, but obviously the GPUs.
50:54
Uh but yesterday I saw a fantastic deal with our our buddy at Lambda uh for you know a big >> busy.
51:02
Yeah, you're busy uh big GPU cluster.
51:04
Is there advisory that you can provide even if CrowdStrike is going to be racking Nvidia GPUs?
51:09
How deep does that partnership go beyond just like cool here are the weights we signed on the line you can use them. Okay. Yeah. >> Yeah.
51:18
I think uh most I mean as I mentioned most of the advancements at this point is research in the harness. Sure.
51:23
>> So we're we're uh our research teams um you know George announced this uh what is it called?
51:28
Cyber super cyber and super intelligence lab.
51:30
super intelligence lab. uh we have a bunch of cyber researchers uh and basically what we're doing is we're constantly publishing where advancements are coming uh in in in the in the ecosystem or in uh these environments like we put out there >> uh a few weeks ago some uh new harness research that we called AVO that showed
51:47
in ARC uh IG uh ARC agi3 uh we could take a a frontier model from 30% accuracy to 100% accuracy >> or agi v3 yeah that's insane >> and that's all open you know research I feel like that's the big surprise over the last month is is people showing what's possible with a different harness versus ARC's standard harness, right? >> Yeah. Yeah. So, so we, you know, share >> Yeah. Yeah.
52:09
So, so we, you know, share all of that open research uh together to advance the industry.
52:13
And ultimately, you know, we're not a cyber company, they're the cyber company.
52:17
Uh they have the domain intelligence, the perception into customer environments to understand like real attack paths that people are trying to exploit.
52:24
trying to exploit. um you add to that these new agentic attack paths uh that people are trying to understand in their environment and and ultimately we you know want to help uh power uh defenders and give them this differential advantage that Jensen and George >> I'll add on to that I think every
52:38
meeting that I have and that that everyone at Crowdstrike has in Nvidia they all start the same end the same how can we help you grow it's the first question it's the last question it's from Jensen all the way to the person at the front desk and so your answer to your question of where like what's on the table everything is everything. Yeah. Yeah.
52:55
>> The whole shop's on the table.
52:55
It's like, what do you need from us? We're here.
52:57
So, like >> when we Who's the best AI partner that we have?
53:01
It's Nvidia because they're helping us take cyber security to a whole new space and we're bringing them to over 100,000 customers and everybody that's on the show floor here today. >> Yeah, that's great.
53:10
Um, how important is uh human design of RL environments for this harness development?
53:18
You obviously have the most insane data collection, correct?
53:23
decades of experience across the entire organization.
53:25
There's a lot of value and ways that I could see if you're there's a huge jump. I'm not surprised.
53:31
I'm sure, you know, congratulations.
53:33
But um but but how much how much is it about actually designing new environments and then go training?
53:39
>> Intelligence, I believe, is really becoming somewhat commoditized.
53:41
I think what's really real in this next chapter of AI is how you contextualize based off of specific situations.
53:48
of specific situations. So the fact that we have Falcon complete data that's managed in detection response data from human analysts that took actions over the last number of years across you know all these different environments the fact that we have frontline uh incident responders that stop the breaches all
54:04
that that data set lets us curate something that's super relevant and super focused and then we take that and we we operationalize it with the harness so that we can bring a better model that's built on Neotron and have a appropriate harness for solving different problems and have an iterative learn learning loop like where this all
54:21
goes in my opinion is you'll see more models and more harnesses from us in the safe mind family that solve different security use case problems and that's all based off of the experience of our practitioners you know crowdstrike is cyber security built by and for cyber security practitioners I think that's really different in the market for us
54:40
versus a lot of the other random companies that you find uh that that say that they're here to work in cyber security like everything is based in solving a real problem >> yeah How are you thinking about educating the customer, the buyer on cost and how to think about the shape of cost in this token maxing? I mean, a
54:56
I mean, a breach can be so devastating.
54:59
Throw all the dollars at it, but at the same time, there's amazing trade-offs that you can do with smaller models, different infrastructure, and different pieces of the puzzle.
55:09
>> Let's start with Justin because I think you've you've been >> helping evangelize open source and trying >> well in our environment. Right.
55:14
So you got to remember so one uh we we are also a big enterprise.
55:17
Uh we we have to look at the same threats as everybody else. Right.
55:21
So and I think a lot of the conversation has been steered around uh code vulnerabilities but in a production environment it's really about also the configurations in your running environment. >> Sure.
55:31
Um and so uh you know for us uh you mean to your point the harness should be able to use uh the best of frontier and the best of open to uh reason through and and figure out uh which problems you you need to use which models for and ultimately uh we assume it's going to be always on.
55:46
Uh we want to start by trying to make it as low cost as possible by providing open intelligence that they can domain adapt.
55:54
Um so the safe mind models are probably the default. Yeah.
55:56
Uh with the exception being the frontier.
55:58
uh if you want to look for very novel new things. Sure.
56:02
>> Um and that kind of gives you the best uh cost uh you know cost benefits when you run this all the time across code binaries and configurations in your environment. >> Yeah.
56:10
We've heard it loud and clear from customers that just going one direction with Frontier Labs is just too cost prohibitive. >> Yeah.
56:17
>> But open source at this point like generic open source is sort of like you don't have the it's it's a compass that's spinning in a circle.
56:22
So what we need to do is have the best use cases, the best results, the best outcomes and also deliver it at the best cost.
56:28
And that sort of is the the aperture that we that we need to play in with this thing.
56:33
And so I think every enterprise is grappling with we have this new line item in COGS that's called tokens.
56:37
Y >> you know 5 years ago didn't exist >> and and it's not like you necessarily say goodbye to anything else by the way like >> you're doing more with your cloud providers.
56:48
>> You're using a lot of software.
56:48
You need to secure all of it with Crowd Strike of course.
56:52
So, you know, I think everybody's in this redistribution or rethink of how you do budgeting in this in this new AI first world. >> Yeah.
56:59
And I think every enterprise is planning to spend more on AI next year, but at the same time trying to be a lot more efficient, right?
57:04
And so that's why these two these closed models, open models can can coexist and actually the industry can continue to thrive. >> Yeah. Yeah.
57:12
Uh, how do you think about the, we were talking about this earlier, but the the the economic warfare between attackers and defenders because the the benchmark performance, all the stats that you mentioned, those are great, but I'm almost more excited about the cost savings because this is a technology that needs to be always on, running all over the place.
57:32
It needs to be uh, you know, too cheap to meter essentially as fast as possible.
57:37
So it can be everywhere because if an attacker is only trying to come through one door, you got to make sure every door is secure.
57:43
So how are you thinking about the economic balance between attackers and defenders?
57:47
>> Well, the way I think about it is price is I mean value creation is always measured economically in a P and a Q. >> Yeah.
57:53
>> What's happening right now is the Q is going out of control.
57:56
>> So like that's the big picture.
57:56
Like there's more attack surface than ever before. >> Yeah.
58:01
>> And that means there's more opportunity to come back to your question.
58:02
There's more opportunity for adversaries to play around. >> Yeah.
58:05
And they don't have to be right every time.
58:08
They just need to be right once. Yeah.
58:09
And go get something off the shelf somewhere >> and use the weapon.
58:12
And if the weapon works, they that's a good day for them.
58:15
>> So do you think 2026 if if we look back in a decade, do you think the attackers are going to be like that was the best year we ever had or do you think it's going to be the moment when the defenders are saying that was the moment we figured things out? >> Going on a limb here.
58:30
Justin can back me up or you have your own opinion, but like Safe Mine changes the curve.
58:33
Like >> I think Frontier AI has disproportionately uh advantaged well one everyone sees advantage. Yeah.
58:44
>> But I think sort of until until like this time it's sort of disproportionately advantaged to adversaries. >> Okay.
58:52
>> And I think it's time to change the tide. Sure.
58:53
>> And that's why we're working together and we want to see that change happen and make that a reality. Yeah.
58:56
Um, there's a lot of great technologies on the floor here.
59:00
There's a lot of great cyber security companies and there's a lot of companies that we're keeping safe all the time, but it's too easy and it's too dangerous for these adversaries to get their hands on things that are way too powerful.
59:09
>> And I think it's a it's another example of a domain where like general intelligence can like come and do attack, but it's going to be pretty expensive to run these attack paths through these like frontier large models.
59:21
Think of it like a like a battleship, right?
59:22
And then what you want to do is you want to help defenders have the equivalent of like drones like super lowcost uh you know models that they can run everywhere so they can run across their entire estate.
59:31
Um and to DB's point like the goal is uh I think to help specialized cyber security defenders have the tools and and their advantage is also they know the code they know the configs that they run the people coming in from the outside don't.
59:46
So on the inside you can do all that recon, you can map your environment, you can uh you know use these lowerc cost models to find uh you know potential uh new attack paths uh and then continue to close them down and do it at a lower cost if you use uh you know these these new platforms like Safe Mind. >> Yeah.
1:00:01
So there's >> you guys you guys are in the position where you can be at the frontier with these capabilities but there's not this like insane pressure from billions of users out there.
1:00:09
Hey, you have to release these cyber capabilities to everyone, right?
1:00:12
Whereas the frontier labs are in a different position where you have hundreds of millions or billions of users that want the best capabilities for things like coding, right?
1:00:22
And and these other capabilities.
1:00:24
So >> well, we have the pressure of lots and lots of big numbers of attack services.
1:00:28
Those are endpoints, identities, cloud workloads, and they're putting a lot of pressure on us because they all need protection.
1:00:35
>> You know, we can't let any of those things get compromised.
1:00:36
So, like the the threat's real, the need is there, the budget's there, but the market's asking for something better and something different.
1:00:44
You don't treat a specialized illness with a generic pill.
1:00:46
You need to have the right the right dose, the right the right therapy.
1:00:50
I think that that's really what we we've done here.
1:00:54
>> And I think with every platform company, we're a platform company.
1:00:56
Uh we know what we are and what we're not.
1:00:58
We're an accelerated computing company.
1:00:59
We're not a cyber security company.
1:01:03
these natural partnerships to allow us to go in a specialized way solve the problem the frontier labs are probably thinking to themselves like where do they really want to uh own call it alpha and go try and compete where do they want to partner
1:01:21
>> um and I think in these areas of like specialized intelligence for cyber security um you know we like generally think the best approach is uh help you know protect critical infrastructure help make the world a safer place and and we'll all be in a better place. >> Yeah. What uh as as a platform company, >> Yeah.
1:01:32
What uh as as a platform company, what is the advantage uh that you see occurring over time around having diversification in the actual chip fleet?
1:01:43
I mean, there's this uh Grock deal coming online.
1:01:45
There's uh there's already a number of different configurations of rack scale servers and all sorts of different back to the gaming chips.
1:01:53
I mean, I see people running AI loads on those, too.
1:01:54
Um what what is the what is the advantage and the shape of that over time?
1:01:59
>> Well, we're an accelerated computing company, right?
1:02:00
So we have to accelerate everything.
1:02:02
Uh and the the reality is um you know different models need different you know capabilities and so ultimately we want to be able to provide the capabilities whether you're doing uh prefill and inference or decode uh have the most uh performant capable architectures.
1:02:17
Um and then ultimately uh you know as Jensen always talks about we're building rack scale infrastructure with seven processors.
1:02:23
Yeah, >> we're trying to be best of breed across all of those so that we can build these large AI factories and drive the best token efficiency per watt um and then ultimately have a great partner ecosystem that can extend that efficiency into these new domains and use cases. >> That's fantastic.
1:02:38
Well, congratulations on the deal.
1:02:39
Thank you so much for coming on the show. >> Pleasure to be here. Thanks guys. Thanks for having me. Thanks for being here. Thanks. >> This was fantastic.
1:02:45
Uh thank you to everyone who's tuned in live from Falcon. >> Great to see you. Cheers. >> Okay.
1:02:52
Um, there are a few more stories that we should get through.
1:02:55
Um, let me tell you about Shopify.
1:02:58
Shopify is the commerce platform that grows with your business, lets you sell in seconds online, in store, on mobile, on social, on marketplaces, and now with AI agents.
1:03:06
Uh, Jordy, was there anything else that you wanted to get to?
1:03:10
>> Give uh John Turnis uh a great follow over on X.
1:03:14
>> Yes, John Turnis has hit the timeline.
1:03:16
Uh, >> what a what a what a moment for X in some ways. Yes. Right.
1:03:19
you know, we're how many years into the platform?
1:03:23
>> Yeah, it still feels like if you assume an important uh position in the world of business, you simply can't afford not to be on X. >> Yeah. No.
1:03:32
Uh the the fact that I mean it's not like he's like, you know, tweeting random stuff or actually like breaking is there yet?
1:03:39
>> Since we started the show, he started ship hosting. >> I'm kidding.
1:03:41
[laughter] >> No, he tweeted hello lowercase very online, very like native.
1:03:46
Uh but uh uh the fact that Jensen's on there, Mark Zuckerberg's on there.
1:03:51
Like the AI conversation is truly happening on X and uh it's exciting to be a part of it.
1:03:55
Uh Apple investors want the new CEO to be an innovator.
1:03:59
This is in the Wall Street Journal.
1:04:01
Ralph Linklair writes um talking about there is one way uh there is there is one way which company observers have said uh has been lacking since the Steve Jobs era rev revving up Apple's innovation engine especially in artificial intelligence.
1:04:15
Tim Cook's brilliance was to take the company Jobs Built and scale it massively.
1:04:19
The year Cook took over, Apple sold 72 million iPhones.
1:04:25
This year, it will be 255 million.
1:04:27
Uh, tripling volumes, hitting annual release dates like clockwork, minimizing risky capital investments and returning more than $1 trillion to shareholders.
1:04:35
That is a size gong moment.
1:04:37
Um, helped Cook multiply Apple's valuation by 13 times.
1:04:40
But uh Ralph Winkler in the Wall Street Journal says that Apple's investors now want Turnis to change things up.
1:04:47
It's not enough to rest on your laurels.
1:04:49
Just focus on operational efficiency. He's got to innovate.
1:04:53
According to the Wall Street Journal, they say uh but in Wall Street Parlayans, the positives look priced in.
1:05:00
Apple's tra Apple stock trades at 33 times next year's earnings compared with the S&P 500's collective multiple of 20 times.
1:05:06
Apple gets that premium even though its earnings are growing half as fast as the market.
1:05:11
Investors are paying up for Apple because it looks safe at a time of broad anxiety over returns to be had on massive investments in AI.
1:05:17
But when the valuation gets stretched, safety isn't safe anymore.
1:05:21
And there are negatives that investors may be overlooking.
1:05:26
In the age of AI, Apple has lost its status as the consumer a as the company that defines how consumers interact with devices, a title it held for 40 years from Apple 2 to the iPhone.
1:05:37
So, uh the Wall Street Journal wants Turnis to take risks, launch new products, uh you know, go go aggressively.
1:05:43
He certainly uh seems like he's stepping back from the Apple Vision Pro sadly, but uh we'll see we'll see what he does.
1:05:48
It'll be a exciting time.
1:05:50
And last but not least, Dyson just released a new AI powered toothbrush with integrated 100,000 pixel macro lens camera for $499.
1:05:57
I know a lot of you people have been asking for AI in your toothbrush.
1:06:01
Uh I certainly know I have.
1:06:04
Uh and uh I'm glad that >> this can't possibly be the first AI toothbrush. There have to be other.
1:06:10
>> Somebody's got to get this uh and try it out.
1:06:13
I I I certainly have never wanted >> What are we doing with this 100,000 pixels >> in my toothbrush or really a camera in my toothbrush, but >> that's not how people measure camera lenses.
1:06:24
They say megapixels, which I think is a million pixels. So, it's actually a 0. 1 megapixel lens camera.
1:06:32
>> But I think we got to give it a shot. >> Okay.
1:06:34
>> Got to give it a shot.
1:06:34
I don't want to judge it too much.
1:06:35
I I do like a good wooden toothbrush.
1:06:39
>> I like a wooden toothbrush personally. Yeah.
1:06:42
>> Uh I think they get the job done, >> but >> oh well, >> we got to try it out.
1:06:45
Uh we'll be back in the Ultra Dome.
1:06:48
>> Yeah, back in the Ultradome tomorrow. >> Cannot wait.
1:06:50
>> We're heading back to Hollywood.
1:06:50
Thank you for tuning in, folks.
1:06:52
It's been an honor and a privilege.
1:06:55
>> We'll see you tomorrow. Goodbye.