theCUBE + NYSE Wired: Physical AI & Robotics Leaders QA2 | Irina Denisenko, Knox

Clips
More from theCUBE + NYSE Wired: Physical AI & Robotics Leaders QA2

Irina Denisenko

CEO

Knox

play_circle_outline Transforming Federal Cloud Security: Knox AI Auditor and the FedRAMP 20x Initiative for Real-Time Compliance and Efficiency

play_circle_outline Infrastructure as code enhances cloud security configuration assessments.

play_circle_outline Transition from manual audits to continuous, AI-driven security monitoring.

play_circle_outline Importance of proactive remediation in cloud security practices.

play_circle_outline Enhancing Data Security Through Cross-Agency Collaboration: Breaking Down Silos for Improved Sharing and Compliance Efforts

play_circle_outline Need for software bill of materials (SBOM) to enhance supply chain security.

play_circle_outline Digital twins and simulation technology are emerging trends in defense investment.

Info
Transcript

Irina Denisenko, Knox

Irina Denisenko

CEO Knox

Irina Denisenko, chief executive officer of Knox Systems Inc., joins theCUBE’s John Furrier and Dave Vellante during theCUBE + NYSE Wired: Robotics & AI Infrastructure Leaders 2025 event to discuss how AI is modernizing GovTech and cloud security. The conversation explores how automation and infrastructure-as-code are reshaping compliance for federal systems.

Denisenko outlines Knox’s approach to continuous monitoring through its AI auditor, reducing reliance on manual audits and streamlining adherence to frameworks such as FedRAMP. With open-source ... Read more

explore Keep Exploring

What is the current state of innovation and cybersecurity in government technology (GovTech)? add

What is the technology used in cloud security and how does it work? add

What is the significance of continuous monitoring in the context of security audits, and how has the approach to monitoring evolved with advancements in technology? add

What measures can be implemented to effectively address vulnerabilities in cybersecurity for federal agencies? add

What is the current status of interagency collaboration and the CISA reauthorization? add

What initiative is the Department of Defense (DoD) pursuing to enhance understanding of software components and supply chain security? add

What are the current advancements and investments in digital twins, synthetic data, and simulations within the AI sector, particularly in relation to government initiatives? add

bolt Powered by CUBE AI

Irina Denisenko, Knox

search

>> Welcome back everyone to theCUBE. We are here in our Palo Alto Studios for the special presentation of robotics and the AI leaders as part of our three days of coverage here kind of midway through the year. I'm John Furrier, your host of theCUBE with Dave Vellante CUBE in-depth coverage with the NYSC, and they're performing the NYSC wired community arenas here, CEO of Knox. She is in the GovTech area. A lot of action. There's a lot of data in the government arena. Great to see you again. Thanks for coming on theCUBE. Appreciate it. You're closing us out for the event. It's our 35th interview.

Irina Denisenko

>> Thank you so much for having me. I'm so glad to be here.

>> Yeah, we're so psyched for you. Now, one of the things that's come up, and I think Dave asked the CEO of VAST Data, you guys have a lot of financial services. He's like, "No, we have life sciences, a lot of government. So there's a lot of data and the data's exposed." And so there's a lot of work going on out there around, okay, how do we be more efficient? We see all see DOGE in the news, but that highlights efficiency and productivity. So we know the government's been on transformation with Cloud. Now AI's here, it seems faster now with AI than Cloud. Cloud was like, "Okay, CIA led it. Other agencies follow." Now you've got the secure regions, secret regions. Now with AI, what's the vibe? What's the sentiment in the government because they got the data and they know how to use data. Certainly some of the agencies do. What's the situation in GovTech?

Irina Denisenko

>> Well, the situation is everyone is aware that they are holding some of the most sensitive data in the world, and yet they need to keep up with the pace of innovation because you know who's keeping up our adversaries? China's using AI, our adversaries are using it. We have to adopt it inside of our DoD and inside of the civilian. And so really the focus that both the DoD and civilian spaces have had in the government is how do we enable us to use the latest and greatest but keep it secure? And so why I'm so excited to be here is FedRAMP, which has obviously been the standard for cybersecurity, for cloud cybersecurity, for software in the government, has been leading an initiative called FedRAMP 20x, which is of course, how do we bring continuous monitoring into the security of cloud, right? Rather than running manual audits of cloud security, how do we do it actually using software, doing it continuously? And so what I'm so excited to announce is that actually on Monday we'll be publishing, but we're breaking the story here today. Our AI auditor Knox AI, which we have been using for many, many months now to actually scan and keep our cloud secure. We're actually bringing it to the entire public sector and showing exactly how it is that we keep our cloud secure by scanning infrastructure's code and by enabling us to in real time know our compliance with frameworks like NIST 800-53, which is really what FedRAMP and the DoD DISA frameworks are based on. And what's even more exciting is we're doing this in partnership with Wiz. We're doing it using tools like Drata. And we're doing this not just for our own technology, but also for Adobe Connect, Adobe Learning Manager and a number of Adobe products, all of which host in our cloud.

Dave Vellante

>> Amazing. So when we talked at RSAC, we had you on theCUBE for a short little hit, theCUBE After Dark. You said the standard up until you Knox was to sort of take snapshots and see how you're doing as opposed to the sort of continuous monitoring.

Irina Denisenko

>> Exactly.

Dave Vellante

>> What's the technology behind that?

Irina Denisenko

>> The technology is actually something that we've been doing in the commercial security space for years, which is infrastructure as code. We all know Terraform, we all know HashiCorp was acquired by IBM. It's an established market, right? DevOps used to be ClickOps and it's sysadmins and things would go down when you'd push to production. We all remember that. Today it's a very different world, infrastructure as code is here. And that is ultimately what unlocks us to be able to know is a cloud configured the right way? When you think about what is cloud security, it's how are you protecting the data at rest and in transit. And how do you know that you're protecting it? Well, what are all of the resources in your cloud environment and are they exposed to the public internet? Are they encrypted, et cetera? Are you using the right levels of cryptography? That is ultimately what infrastructure as code allows you to do. And so what we've built, again, in addition to using tools like Wiz and Drata and so on, is we are scanning the entire boundary, the entire cloud environment at every resource level using infrastructure as code to understand the configuration of every resource to understand if it is secure. And what's even more exciting is at RSAC we announced that we FedRAMPed Spacelift. Spacelift forked Terraform famously a year plus ago. But famously, particularly since the HashiCorp BSL has come in place, Spacelift has seen an explosion in usage because they are the true open source infrastructure as code provider. Well, now not only are they the only true open source Terraform provider, they're the only FedRAMPed solution in the market. And so we're so proud and so honored to have been able to be the folks that enabled them to achieve FedRAMP. It also means that not just Knox can scan its infrastructure's code boundary. Anyone running a federal boundary can scan that infrastructure's code and can understand their compliance. Which again, coming back to why is this all important, I think we can all just fundamentally kind of with common sense appreciate why taking a snapshot once every three years of your cloud environment is not the way to ensure security. You need to be in real time, you need to be in run time. And that's exactly what technologies like this enable-

Dave Vellante

>> How you doing? Not too good. We've been exposed for three years.

Irina Denisenko

>> Right, right. It's like you're finding out.

Dave Vellante

>> How many days ago?

Irina Denisenko

>> Yeah, 900 days ago I was exposed. I mean it's not-

>> So you got to speak cloud native or DevOps and infrastructure code to do it with an AI system and agents and whatnot I imagine is more there. So that's key. So explain how that works and then talk about the alternative. What's the other option? Staff, expenses probably going to be there, time. So you got exposure on time, just chasing down security audit things.

Irina Denisenko

>> Absolutely.

>> Scope that.

Irina Denisenko

>> It's kind of tail as old as time, right? Manual audits where you are sitting in a room with your auditors or even just as your own team, right? Continuous monitoring is the name of the game in the FedRAMP space. It's also the name of the game in security, right? But continuous monitoring, what does that really mean? Is it eyes on glass by hundreds, thousands of employees on your staff and the employees of the tools that you buy for your staff? Or are we moving now into this modern era where AI agents are able to understand what is happening and make decisions based on that? And again, we are just one of hundreds of companies that are bringing these types of tools to market. AWS re:Inforce was earlier this week and we saw a number of announcements both from AWS but as well as Trellix and just Orca. A ton of companies made-

>> How does it work? Do you train off of a DevOps speaker? MLM that's DevOps centric? What's the learnings from?

Irina Denisenko

>> Absolutely. So first and foremost, our cloud, the Knox Cloud has been running for 10 years straight. What does that mean? We have 10 years of audit data. FedRAMP is only about 13 years old, so there is really no one out there who has as much data specifically on a federal cloud. We have all the revisions. We've gone through Logj4, we've gone through all of it. And so first of all, there's a ton of proprietary data. Second in our vector database sits every single artifact of Terraform code that exists on the public internet. And anyone can pull that down, certainly. We just took the initiative and did it. And so it's something like 70,000 or a 100,000 kind of files if we were-

>> So you know the code inside and out?

Irina Denisenko

>> It is literally every ... What's beautiful about the Terraform and really the infrastructure's code movement, it's not only Terraform but it's primarily is it's all open source. It's on the public internet. And so you can pull it down and you can train on it. But again, when you combine that with our proprietary data of running this cloud for the last 10 years-

>> Pure baseline code, it's baseline. And you look at that's now input?

Irina Denisenko

>> Absolutely. Right. And you can see, okay, if something is not compliant, how can you then make it compliant? It's not just about detecting where are you out of compliance. It's also allowing you to say, how will you remediate? And also to be able to assess in context. What does assessing in context mean? Well, you might have, the example I love to use is you might have port 80 open to the public internet. What does that mean? Well, that's a no-no as far as if you just are by the book by NIST 800-63. Well, we at Knox run a ton of web conferencing tools in our cloud. Adobe Connect is one of the tools that we run in our cloud. That's a web RTC-based video conferencing tool. If you close down port 80, the entire tool shuts down. You cannot do that. So what do you do instead? Well, you don't shut it down. You put mitigating factors in place like you put a WAF in place, a firewall, you can put other things in place. The point is, it's important to be intelligent in how you remediate. And again, being able to do this continuously quickly at the time that vulnerabilities are detected is critical. And again, really the way we really look at this is we're really just providing federal agencies, FedRAMP itself, the DoD tools that the commercial world has been using forever. We've been using forever. Now we're really bringing this into the hands of the folks frankly who need it most because they are the ones who are being pushed the hardest to adopt some of these most cutting-edge technologies.

Dave Vellante

>> I love this conversation. You were at re:Inforce, we were both at re:Inforce this week. John, you remember in 2019 the first re:Inforce-

>> Yep, we were there, Boston.

Dave Vellante

>> In Boston and we had theCUBE there. Amazon is uniquely optimistic about security. Remember Steve Schmidt? He was like, "All this negative talk about security being broken-"

>> And do-over....

Dave Vellante

>> "That's wrong. Our security's great and so-"

>> We're complex, therefore we're secure.

Dave Vellante

>> So I wrote a piece and the reason I love this is so at the time we were like, okay, the cloud became the first line of defense, but then you have the shared responsibility model, which there was a lot of education on that. That was a stopgap at 2019 re:Inforce and people, many in the audience, oh, shared responsibility?

>> Yeah, you own the problem, not me.

Dave Vellante

>> I thought my security was good. So that was sort of educational. But then as things evolved, I felt like they were doing a great job with their infrastructure, but still putting a lot on the CISO and the app dev teams. So now when you think bring infrastructure as code and the whole shift left movement has helped. My post this week was on AWS. The title was re:Inforce 2025 Elite Security, Still Too Complex. You are filling that void and simplifying some of that complexity for that shared responsibility model. Is that a right understanding?

Irina Denisenko

>> It's a great point. When you think about what Knox really is, we are the largest federal managed cloud in the world. What is a managed cloud? We take all of that responsibility. It's not just shifting left, it's just give me all of the-

>> Your shift.

Irina Denisenko

>> It's just I take all of that, right. That's right. And that's a huge responsibility that we don't take that lightly. That's why we built an entire suite of AI agents in order to just ... And by the way, not only we built our own, we are customers of Wiz, we are customers of CrowdStrike, we are customers of Tenable. We literally buy everything best in class to secure this cloud. Make no mistake. But at the end of the day, the reason our customers work with us and trust us is because we are taking on the entire burden of the FedRAMP compliance aspects of this. And over time, we are seeing customers push actually more and more workloads to us because they see us do a great job in serving and hosting the production environment for their federal customers. And then they say, "Well, what about my regulated industry customers, the banks I serve, the healthcare institutions, I serve the Fortune 100 that I serve? Can you hold those workloads as well?" And we're seeing more and more of that. So indeed security is still very complex and there is absolutely a desire. We see it in the market for I would say mid-sized, even some enterprise companies to say, you know what? Let me just pass the buck and it's going to be someone else's problem. And by the way, I mean that's why even AWS, right? I sat down with Hart Rossman who oversees the entire managed cloud division of AWS. You look at Mandiant at Google, managed cloud is not a new concept. Folks want to shift that responsibility where it's new and where we are really pioneering is in the federal space, in the harshest environments, in the most secure environments.

>> And more demanding and they're under-resourced.

Irina Denisenko

>> Absolutely.

>> And they want to have a reliable partner.

Irina Denisenko

>> Exactly. And especially, again, going back to where we started all of this, especially in the age of we need to adopt AI. And you know what AI isn't going to do? It's not going to sit on-prem. It's not going to sit air-gapped. Right? To your point, about 2019 was still felt perhaps like early days of cloud, even though by no means was it. Cloud is only going to accelerate with AI because it has to be internet connected in order to achieve its full potential, which means right now, what is it about 20% of all workloads are run in the cloud and 80% still is on that-

>> In government tech I'd say that that seems to be, but enterprise is still in an on-prem because they're nervous.

Irina Denisenko

>> Exactly. Exactly.

>> But they still have cloud storage.

Irina Denisenko

>> Absolutely.

>> It's all one thing. It's distributed computing.

Irina Denisenko

>> Correct.

>> As we see. Well, the question I have for you, I know we were talking before you came on camera, you do a lot of salon dinners and meetings with the top agencies and the top CISOs. I know you really can't name names, but just give us a feel for the conversations that are going on. How alert are they to the opportunity and the risks and the requisites required to onboard AI in a meaningful way?

Irina Denisenko

>> Absolutely.

>> What are they talking about? What are some of the conversations?

Irina Denisenko

>> Yeah. Well, look, again, not to name names, but we just held one of these last week in conjunction with AWS Federal and Gartner Cyber and Coalfire RAMPCon. And long story short, the topic of the moment is only in the last 60 days have we had confirmed workloads and confirmed government data touch LLMs that are internet connected. Prior to this spring 2025, it was fully air-gapped, it was IL-6, it was fully on-prem, it was a different type of deployment. We've now crossed the Rubicon, and so we're here. And so really the topic of the moment is are we secure? How are we protecting against prompt injection and so on? And so the types of folks that we have at the table at these are folks from the top LLM and foundational model companies, the folks from the top cybersecurity companies who are really in many ways taking the responsibility to ensure that we do stay secure. And then of course, the owners of the data themselves, the CISOs, the CIOs, chief AI officers of every federal agency. It's interesting, a year ago there was no chief AI officer and now every single agency has a chief AI officer because of this exact issue. So that's the topic of the day. I'm sure your next question is, so what are we doing and are we secure? And really the answer is the cat's out of the bag. And I personally have not heard a single definitive, rest assured we've got it type of answer. No one's been able to give that.

>> That was one of my questions. The other question I was going to ask you on this chief AI officer is it's usually the CTO or CISO's extra job that they have to do because there's a lot of technical acumen needed. It's not like a political position of, "Hey, let's herd the cats and get a strategy." It's kind of a security calls to arms. It's not so much what's our AI strategy? What's our AI strategy and what's the security posture look like?

Irina Denisenko

>> That's exactly right. What workloads can you unlock? What missions can you serve? But how do you serve them securely? Because all of these require the proprietary data that only these agencies have, whether it's the DHS or the Department of Energy or really name any agency within the DoD, right? They have a ton of proprietary data, a lot of top secret data that has to remain secure.

>> Is there a movement in the DC circles? I'm just not up to date, so more of is a current question for you is how current is this customer-centered excellence concepts that we see in big hyper neoclouds and enterprises where in a big bank or something there'd be a department. It's more of an organization to get people up to speed. Agencies have been siloed and have their own thing. How much of this center of excellent cross-pollination is there across agencies? Does it exist or is it more by siloed agency group or doesn't even exist?

Irina Denisenko

>> Yes. The answer is yes, all of the above. Look, I think agencies really hats off to them for doing everything they can to, again, at the end of the day, the buck stops with them for their data. That said, there's a lot of interagency collaboration on many topics like FedRAMP for example. And then you have agencies whose job it is to sit across all. CISA, for example, is one, and we're actually right now, one of the things that I was engaging with both with the Senate and with the House of Representatives last week is the CISA reauthorization, right? That's led by Senator Rand Paul, that committee, and we're right now going to the mat on the reauthorization.

>> So at the Amazon Web Services DC Summit, we were at . You guys were at the opinion force together, but I was at the DC Summit, which used to be called the Public Sector Summit. And the conversation there was, besides some of the Doge things and efficiencies and using agents, the conversation of horizontal scale across agency came up. So the old school, not my budget, not my department, is kind of breaking down. So you're starting to see new behavior culture. And so I was curious on that point about the center of excellence, but there's real mindshare that agents will talk to other agents, whether it's NPC servers, you have different states, stateless, whatever form you use. How much aware of the agencies on collaborating? I mean security they collaborate all the time, but this is real time. The agents will interface with other agents and share data. DoD might want to talk to the CIA or NSA.

Irina Denisenko

>> Right. Well, look, I think we're seeing, first of all, a one-government movement from a procurement perspective, and we're seeing the GSA and the OMB really take control of that so that it is really one budget and not everyone is siloed. The other thing that we're seeing is very specific use cases of this cross-agency data. We've seen a lot in the news about what Palantir's doing with bringing data together and building kind of a one-data set. I think more to come on that from what they're able to share, but indeed, we are trying to break down data silos in particular because for the first time ever, we can actually parse through that data and truly provide insights across them. MCP is a huge accelerant of that, right? I think at this point there isn't a tech company, certainly not a cutting-edge tech company that hasn't leveraged MCPs. We leverage it. In fact, in our AI auditor, how do we connect into tools like Wiz, like Drata? We use their MCP servers or we build our own to connect to their API and it accelerates the rate of integration. It goes from weeks to hours. That's huge. And by the way, lowers the barrier of how technical someone needs to be in order to plug into that tool app, whatever it is you're trying to plug into, because MCP is really this, it's like API integration, but an AI agent is doing it, right? I mean, that's what it is.

>> It's really a bright spot in this year's organic defacto movement. It galvanizes everyone around the common goal of can we talk to each other?

Irina Denisenko

>> Absolutely. Absolutely. It literally breaks down data silos. Now, the flip side of it is, well, how do you ensure that this AI agent isn't just pulling in all the data? I'll give you a very specific example. We plug into GRC tools like Drata and Vanta and so on through MCP servers. Well, those tools are plugging into HR systems. In an HR system for the goal of FedRAMP compliance, I might want to know that an individual that was terminated from a company that their laptop was received back by the business, by the company, because that's a security compliance matter. I certainly don't want to know why they were terminated. That has nothing to do with FedRAMP. The challenge with MCP agents right now and servers is that there's not a whole lot of where do you stop? They're kind of a, "Let me go get all the data at the same time."

>> It's the hunting license.

Irina Denisenko

>> It's complete free for all. And where we're seeing a lot of great work. Anthropic is doing this work, certainly all of the foundational models, but even folks like Red Panda is doing this and a number of others. They're building really interesting tools to guardrail these MCPs, and we're leveraging a lot of that. We're certainly building our own. The point is it's everyone's job to ensure-

>> And there's also state requirements having that state of the data.

Irina Denisenko

>> That's exactly right. That's exactly right. So it really is early days, and again, that means a lot of opportunity, but it also means a lot of risk.

Dave Vellante

>> Well, and that's where A28 presumably comes in, right?

Irina Denisenko

>> Sure.

Dave Vellante

>> You talk about state. I want to ask you about MSSPs because that market is exploding, especially for organizations that just don't have the skills, which is a continuous problem in security. You've mentioned a couple of partners. Are MSSPs candidates for partnering with you or do they see what you're doing as their domain?

Irina Denisenko

>> It's an interesting question. Look, I think first of all, there is a huge explosion of just tech-enabled tooling that is going on right now. In the same vein that you see MSSPs exploding, the other thing you see exploding is, well, frankly, vibe coding, low-code no-code, vibe coding. Workloads that used to be completely manual or that folks didn't even think that they could touch are now being automated away with just apps that someone can spin up. And so to go back to your question, there is just too big of a TAM for us to be really competing. And where we're really unlocking value is really for companies that they're scaled. I mean, they're 300, 500 million in revenue. They are leveraging MSSP's, but they also have a huge staff where they've found a lot of resistance and just frankly, inability to succeed is to achieve FedRAMP authorization because they cannot get an agency sponsor or frankly, they can't get the resources to build a FedRAMP compliant cloud. So if anything, we coexist. Certainly there's a number of conversations in play around partnering, and again, where I really see the MSSP's is really exploding in the best way is at that mid-market.

Dave Vellante

>> Right. And I would just think it would be a reasonable channel for you.

Irina Denisenko

>> Absolutely, absolutely. And particularly as the mid-market looks to scale into federal regulated industries, etc. It's a natural play.

>> What's the biggest challenge you have going forward now? You got a great growth opportunity. The MCP A2A is good. These are good protocols, bring things together. As the government expands, what's your forecast for what will happen next? Will we see the brakes pulled a little bit on AI to get nailed? The security posture, innovation equation. Is it full throttle pedal to the metal cloud migrations with AI?

Irina Denisenko

>> Look, I think at the end of the day, even though we're at this incredible moment in technology, you still have change management and you still have natural cycles by which folks need to get used to new ways to do things. And so at the end of the day, the challenge is always making sure you don't go so fast that you break a lot of things. And we certainly see a number of very measured, I think very smart approaches, whether it's at the DHS, whether it's within the Air Force, the Marines, I mean, I could go on and on of very specific examples where folks are implementing cutting edge technologies, but they're doing so in a proof of concept. Then they're building it out and building it out. And so the forecast is it's not going to be overnight. The government is not going to move as fast as perhaps some of the, I'd call it more well-known venture backed companies, obviously, but they will move faster and they are moving faster than I think anyone has ever seen government agencies move before.

>> Irina, the final question for you is obviously they are going faster. We're seeing some evidence. It's a good sign and more tech people are getting involved in the law making side, which opens up a little bit more aperture to lawyer to tech personnel making these decisions. This is the Robotics AI Infrastructure Leaders Conference, so I have to ask, military has to look at the physical aspect of DoD planes. You're seeing drones, you seeing what's happening in Israel. Technology and robotics will be a huge part of tactical edge, execution and also old school assets. Software comes in, you've got more robotics. What's the DC scuttlebutt or public information you can share or story around the thinking around, okay, we have to upgrade our military. We want to have all of our government services modernize. Where does robotics and AI infrastructure trends that we have now connect into some of the thinking around ideas around better service, cost-effective, reliable, safe? What's your thoughts on that?

Irina Denisenko

>> The thing that I'll point to is the initiative called SWIFT from the Katie Arrington and the DoD CIO and CISO offices. And really where they're pushing and where we're very excited, we participated in the request for information about a month ago. They are pushing for, on the software side, a software bill of materials, SBOM, they're pushing very hard into that in addition to SAS, DAS vulnerability scans and so on. The point is we want to understand what's in it in the same way that, to your point, there is a physical aspect to everything the DoD does or many things the DoD does. There, a bill of materials has been the name of the game and supply chain intelligence has been the name of the game for a long time, and the ante on that is only being increased, right? We've seen, for example, over in a previous life, I was in the drone technology world and very familiar with DJI and so on. We've seen obviously a massive pullback from DJI and other Chinese manufactured hardware within the DoD in particular. I think we're only going to see continued trend there, but that means that there's going to be more investment. There has to be more investment in US and Allied nation technologies because frankly, when you pull away some of the, frankly, the best technologies which are coming out of China and you don't have something to replace it with, that's not good either. And so the-

>> There's gaps there potentially, evaluate the gaps.

Irina Denisenko

>> Absolutely.

>> Yeah. And also, not to get all nerdy, but the fun stuff in the AI factories physical AI world is digital twins, synthetic data and actually doing the simulations, just progress in where they are on leveraging digital twins, simulation. How advanced are we on the government side there?

Irina Denisenko

>> Well, I will just point to the fact that Meta just acquired scale, and I think that-

>> Invested.

Irina Denisenko

>> That's true.

>> Invested. Invested .

Irina Denisenko

>> You're right.

>> Technically invested.

Irina Denisenko

>> You're right, but I think it speaks volumes to the fact that Alexander is joining really the AI leadership team at Meta officially, and when you think about what scale was able to do on digital labeling and really enabling all of the LLMs to have all that training data, the next frontier is physical and scale has been very involved with the DoD for many years already. I think, again, this really points to the fact that when you look at the now greater investment in what they're able to do, you pair that with Meta in particular has obviously made a ton of investment in glasses in the augmented reality world. Obviously with Oculus as well. You see what obviously is doing and so on. The point is there is a ton of investment and what's really nice is it's coming into companies that have had a lot of experience with the US government already.

>> And the open source Llama will help your AI.

Irina Denisenko

>> Certainly. Certainly the open source is always a double-edged sword with the DoD in particular, obviously, and so you have to be careful there.

>> Not to have the baseline with the AI tools you're scanning.

Irina Denisenko

>> That's right.

>> You have the code.

Irina Denisenko

>> That's right. And it's a little bit of a cycle because you're using the tools to scan the tools that you are using.

>> It's a culture change for sure. I'm sure the DoD, but I like the energy. Thank you for closing us out here on our day here. Thanks for sharing everything.

Dave Vellante

>> You're a great guest. Very impressive. It's great to have you back.

>> All right.

Irina Denisenko

>> Likewise.

>> CUBE AI just got a lot smarter.

Irina Denisenko

>> Yeah. Thank you.

>> Better roll. Thanks so much. This is theCUBE here for our coverage of the Robotics AI Leader Summit, theCUBE plus the NYSE partnership. In collaboration, NYSE Wires is booming. It's an open community where people are sharing. I'm John Furrier with Dave Vellante. Thanks for watching.