AJ Shipley, Crowdstrike | Crowdstrike Fal.Con 2026

Good afternoon everyone and welcome back to the cub's live coverage of Falcon 2026 here at the Mandalay Bay in Las Vegas. I'm your host Rebecca Knight alongside Dave Volante, the co-CEO of the Cube. >> Hey folks. >> Hello. And I would like to welcome to the show for the first time ever. Okay. First time. Yes. AJ Shipley, chief product officer at Crowdstrike. Thanks so much for coming on. >> Thank you so much. So, you're a first- timer here at Falcon, but you are a cyber veteran and you've worked at some of the the biggest names in technology. Wh what drew you to CrowdStrike? What did you come here to build specifically? >> Well, I think, you know, for as long as I've been in the industry, um, and for as long as CrowdStrike's been around, we all kind of looked at CrowdStrike, you know, maybe a little envious because they were the premier cyber security company and have been the premier cyber security company for so long. uh and I was always drawn to the mission and and and you know of cyber in general. Their mission of you know stopping breaches um always kind of you know spoke to something inside of me and you know in parallel I've always felt like you know running product uh is the best job in the world at a product company. Um it's it's so much fun. And so when you have the opportunity, candidly, to have the top product job um at the premier cyber security company in the world, that's one of those things that you know, you don't run towards. You kind of drop everything you can, steal the fastest car, and drive as fast as you can. So yeah, it uh and it's been an amazing ride so far. But uh yeah, that's what it is. And it's a great great group of folks. So it's easy to see why they've been so successful. >> So you kind of led I'm going through the announcements this week. Yeah, >> I was on the the embargo the pregame that you largely well you were a big actor there so it was key. Well, let's start with Guardian. >> Um there's a little bit of Pangia in there I know and there's also some Crowdstrike DNA you know some some organic stuff. >> Um tell us what problems does it solve? How does it work? >> You know where's it fit in the whole portfolio? Yeah. So when we when we bought Pangia and really the the genesis of Guardian being Pangia, you know, the the primary attack surface was or at least what we were trying to monitor was a human sitting at a keyboard putting a prompt into a browser in order to go to a a chat agent and get a response back. >> But agents don't type. >> But agents don't type, right? And you know back in the spring a lot of us were using um you know chat GPT and you know open clause came along and then we started to see this proliferation of of agents down on the endpoint. And so first and foremost we needed to get that visibility into all of the agents that were you know kind of operating in an environment. I sort of equated to the early years of the cloud access um security broker uh market, the CASBY market where you know first and foremost you just needed visibility of all of the shadow SAS and I think what a lot of organizations need today is just visibility of all of the shadow IT and the only way that you get that is to be where the where the I'm sorry visibility of all the shadow AI the only way that you get that is to be where the AI is running which is on the endpoint and so what we did there was we took a lot of the the organic you know kind of capabilities from pangia pushed it down on the endpoint to get the visibility and then provide that visibility to a policy maker so that they could decide what was sanctioned, what was not sanctioned and then start to assign some control around these these uh agents. >> Okay. And then Aentic IDP you guys announced that today I think Y right? >> Um I think you said uh there on the pre-all there's like 100x more agents than than humans. >> I'm hearing 90. Let's let's round up to 100 because it's going it's going to 200. Um and you basically talked about how the legacy ID security model is is a breach waiting to happen is what you said. This comes out of the Signal Yeah. acquisition. Okay. So how do you uh apply identity to to agents? Let's start there. >> Yeah. So it starts with signal really what Signal did so well and why we acquired them is it eliminated this idea of standing privileges, right? So when you come into an organization you're given an identity, you're given um certain you know privileges to access things. Over time those privileges expand. Um very rarely do they ever get pruned but the point being is that that identity is kind of established you know on day one and not necessarily kept up to date. We've always tried to you know talk about just in time provisioning from a zero trust perspective but it was hard under the best of circumstances with kind of traditional identity access management solutions am solutions or IDPS. Now agents that are getting spun up and spun down in real time at runtime just compound that problem exponentially. And so Signal's approach to zero standing privileges and just in time provisioning of the accesses that somebody or an agent needs to do their job was the core capability that we extended. The way that it works is it assigns a unique identifier to an agent. You might have 90 or 100 of those agents that you know got spun up that day or that morning. But with that unique identifier, we can then create privileges through tokens and pass those tokens around and only give the agents the accesses to the systems that they need in order to do their job it at runtime and then revoke those privileges as soon as that task is accomplished. And that is hard if not impossible to do with a legacy IM or IDP solution. >> Okay, I think I get it. Uh because the question I wanted to ask that we ran out of time in the pre-call the the the press and analyst embargo call was >> can you walk us through how agentic IDP limits blast radius exposure because that's something that you mentioned was a benefit and just kind of double click on that a little bit if you >> yeah well first and foremost it's this idea of zero standing privileges which you know you can you can also think of from a zero trust perspective which means you know we are only going to give an agent the privileges and the accesses to the specific system that it needs in order to do its job. So let me give you an example. Maybe Becca in finance goes to Claude and says, "Hey, I want you to pull me a report or build me a report." And for one reason or another in that prompt, the agent thinks that it needs to go off and hit a code repository that the developers are using in order to check code in. Becca and finance obviously should never be accessing a code repository, but the agent doesn't know that. The agent's just going to try to accomplish its task. And these agents reason in a way where it might think that that is a completely reasonable thing to do. So with our agentic IDP, we can first assign an identity to it and then we can give the privileges to the agent first and foremost that don't exceed what Becca's privileges should be. Like an agent that Becca spins up should probably never have more privileges than her. It should probably have a lot fewer privileges than her. And with zero standing privileges, we can give just the very granular access to that agent to do that job based on Becca's context, who she is, what she traditionally accesses in an environment, what she should have access to, and not give those agents any more accesses than that. And when you can't always tell what an agent is doing, which is one of the problems that Guardian solves by being able to inspect all the the prompts, they're doing so much stuff all the time that when it's very difficult to tell what it's doing, you want to be able to give it the minimum amount or least privilege possible. And that there limits the blast radius because if that agent does something it shouldn't be doing or gets compromised for some reason, it only has accesses to the bare minimum sets of things that it should. Okay, so that's super helpful. Follow-up question if I may. Uh reading through the anatomy of the hugging face hack. Can can Becca's agent, you know, collaborate with some other agent that might have that that that access? How do you you know what part of the the Falcon platform sort of detects that and how do you remediate that? >> Yeah. Well, so the Falcon platform first and foremost gives visibility into all of these agents that are spinning up, right? And so, you know, if you take hugging face and you know the,000 or,200 or,400 agents, right, that we're spinning up that we're kind of swarming to the problem. Guardian first and foremost is going to give us all of that visibility to all of those agents spinning up in real time. Then through our agentic IDP, we can assign a unique identifier to all of those agents that are spinning up. And then with the just in time provisioning and zero standing privileges, now all 1,400 of those agents we know about, they have unique identifiers. And we can create unique policies for each one of those agents so that they are not able to collaborate and work together to access systems together in collaboration if they shouldn't be having those privileges to begin with. >> Okay. So, so I'm presuming you're not saying there's a zero probability that that could happen. So, unless maybe you are if you are you know better than going to say that. So, okay. So, there is some probability, right? And there's some unknown unknowns. >> So, is there a kill switch in the system that I can hit or that my agent can hit in case something goes ary? >> Uh, there's not a kill switch in the system right now, at least in the ways that I would think about a kill switch. Um certainly something to think about. I'll take it back as a as a good product person as an idea and say like is there a there there? >> Okay. Um I mean I don't know how that would work. >> Right. That's that's where my head was like how would that work exactly? >> And who's who's who has their finger on that? >> Yeah. And does it require two keys get get inserted >> into how fast can those two keys occur because a lot of damage whoops can you know propagate very very quickly. Yeah. >> Interesting. But I think look if we if we limit the access if we if we remove all standing privileges right and we just assign access at the last minute to only the systems that they need again I think I think the potential is there again leveraging AI to help inform those decisions as well the potential is there to to to reduce down the attack surface or restrict the the accesses such that the type of damage that we saw in a hugging face attack or or an anthropic you know attack is far less than before those sorts of capabilities were in place pregardian. >> I mean, I got to be straight with you. I'm such a Rebecca knows this. I'm such a crowdstrike bull, but I'm also an independent analyst, right? And you know, we're independent media. So, I'm like, "All right, I got to find something >> to pick at." and and so but but I think it's a legitimate concern that maybe a lot of people aren't thinking about which is as I get more telemetry as I get more data I'm I'm giving more autonomy to Falcon that means that the trust bar rises dramatically >> absolutely >> and and so who's accountable and I and my answer is you the customer are ultimately accountable for your own sovereignty not not like territorial sovereignty but there's a shared responsibility model which is kind of fuzzy right now and I think we're working through that and that's why I'm sort of >> raising these questions hopefully for the benefit of our you know customer audience that needs to think about there's there's always something new >> yeah and new risks that are emerging and I don't know how much >> conver look it is shared responsibility we have a responsibility as a vendor but the customer also has a responsibility um in order to help provide context that the vendor we as the vendor can use in order to make better decisions I'll give you an example um only uh a customer would know what the criticality of an asset in their environment is. Right. >> Right. But we should be able to take that criticality and use it to inform the decision-m processes or or inform the agent so that they are better to make a recommendation on potential remediation activities. Um and that only comes through that shared responsibility model. >> That's right. We had a a guest on yesterday who was actually that was their business. They're a partner of yours. I apologize. I can't remember exactly who it was, but they would help, you know, prioritize risk, assess risk, so you could fence that off or or approach that with, you know, differently with, yeah, greater controls. Yeah. >> One thing that's been so clear at this Falcon conference is is the ecosystem and how important it is and how much you're doing with people who are your organizations who are your competitors in other contexts. >> And and it's also clear that cyber is a team sport. No one no one organization can do it alone. How as the chief product officer do you decide what you're going to build internally versus how you're going to look to fill in the gaps with other partners and other maybe smaller companies? >> Yeah. Well, I think great question. um first and and one of the things I love about CrowdStrike is is again just this this deep sense of purpose to you know kind of enabling the defenders with the tools that they need in order to be able to ensure that our society is more secure from a digital perspective. Anyways, I'll just kind of as a segue, um, something that I've always shared is, you know, it's kind of our responsibility and our privilege candidly to protect the loved ones that we all care about because candidly we are all customers of our customers. And my mom's, you know, kind of mortgage is held by a large customer of CrowdStrike. My brother's, you know, and their family's prescriptions, you know, are at a large customer of Crowd Strikes. And if those customers get compromised, then potentially my mom or my brother's, you know, mortgages or prescriptions get compromised. Um, and that's sort of like the overarching mission. And so we use that as kind of a limus test of how are we best able to deliver security to protect the community or the crowd, right? Where we are all customers of our customers, but at the end of the day, we're all, you know, operating in a society and we want to make sure that we are as safe and secure as possible. Now, there's going to be things in there that CrowdStrike is probably uniquely suited to do. And then there's going to be portions of the attack surface candidly that CrowdStrike needs to partner with folks to be able to cover because it's not native in our portfolio. But at the end of the day, what is true and what has been true throughout my entire career is that the adversary doesn't care whether it's one vendor or multiple vendors. The adversary is going to, you know, find the the the weak link in the armor or the weak link in the chain there and do everything that they can to exploit that. And if we are not, you know, coming at it from a community perspective and the power of the crowd, whether they are competitors or or not, if we're not coming at it from that community perspective to cover all facets of the attack surface, to cover all of the assets in an environment that an adversary is after, then ultimately it's going to be my mom or my brother or your family members, you know, who are compromised. And that is probably the overarching litmus test of why we do what we do. >> The real mission. >> I got to ask you this question. It's kind of the there's there's two main companies that are pursuing this consolidation strategy. You used to work for one, now you work for another. Different philosophies, but going after the same thing. You got endpoint, identity, cloud, nextgen SIM, AI, security. These are coming together. I'm interested in in your response to the common architectural layer that makes >> Falcon, you know, one platform. It's not like Paulo a collection of these sort of acquired adjacent products. But to credit to them, I mean they they're an engineering company. They do do good work, but different philosophies. You've never had to rewrite the platform, right? So, but I I want to understand what is that common architectural layer and from your bias standpoint, why is the crowd strike approach more attractive for customers? >> Uh well, the common architectural layer is the falcon sensor, right? And the falcon sensor from day one has been a single sensor. It sits on some of the most valuable real estate in the world. Uh we talk about the endpoint but the endpoint the endpoint can be a container running up in the cloud. It can be a traditional endpoint. It can be an embedded network device. >> Could be identity. >> Yeah, kind of. No, >> I don't think about that. Um I I think we certainly think about identity as an asset. At least I think about it as an asset. I don't I don't know if I think about it as a device. Fair enough. But where I was going with that is um you know I think our architectural approach and our differentiation is that is that single sensor that that lightweight sensor that collects telemetry um and allows us to be able to apply our analytics on it in order to detect maliciousness and then respond to it and then take those learnings and feed it back in so that hopefully you never get compromised a second time the same way that you did the first time. Um, I I can't I can't overstate how important that that that concept of a single sensor lightweight agent is because because anything that you can do to reduce complexity, you're going to improve your security. Um, and then what I would say is probably unique or differentiated about CrowdStrike, sure there there's a sensor and there there's the the the platform and all of the techn technological innovations, but it really comes down to the people. And what I mean by that is especially in an age of AI, you know, just said we are a net data producer, right? Or net data creator, but also at the end of the day that data comes from the industry's arguably largest managed detection and response organization in Falcon Complete that every single day is securing tens of thousands of organizations detecting and responding to adversaries. It comes from the industry's premier, you know, uh, counter adversary operations team, cyber intelligence team that, you know, probably knows about the adversary better than anybody else out there and is constantly evaluating what the adversary is doing, their tactics, techniques, and procedures. And then, you know, the industry's leading incident response organization that gets called in anytime that there's a breach, right? And those people and that data and that learning is what informs the models that we're building as well as the detections that we've created in the platform and ultimately the defenses that we've created and that data is candidly I think the differentiation that goes into and leverages that single sensor that that that that platform advantage that we have in order to make sure that we're delivering the best. >> So it starts with that single and it's a lightweight sensor is my understanding >> and you got a big install base now. So that you know that's the flywheel feeds the telemetry. Why is that sensor so hard to copy paste in a in such a copycat industry? >> Well I mean I would leave that question probably to the the engineers who built it the original you know Alex Yescu and and and all and Hyasin and and all of those folks who architected and built it. They could probably answer that question far better than I can. What I know is that, you know, as a product person, I'm going to I'm going to leverage every advantage I have and keep it and that is an absolute advantage and so I'm going to I'm going to use it to my fullest advantage. >> Great answer, AJ. A pleasure having you on the cube. Thank you so much. >> Thank you so much. I enjoyed it. >> Thanks. >> I'm Rebecca Knight for Dave Volante. Stay tuned for more of the cub's live coverage of Falcon 2026. You're watching the Cube, the leader in enterprise tech news and analysis.