Latest / Expert Insights Podcast / #71. Setting Copilots up for Success (Ben Kliger, Zenity)
Transcript
- 0:03SPEAKER_01: Hello and welcome to the Expert Insights Podcast,
- 0:05SPEAKER_01: your go-to source for cybersecurity insights from industry experts.
- 0:09SPEAKER_01: I'm your host, Joel Wits.
- 0:14SPEAKER_00: AI has a lot to offer in terms of productivity benefits.
- 0:18SPEAKER_00: It's able to sift through a superhuman workload.
- 0:21We're seeing the rise of co-pilots helping users ask questions
- 0:24and navigate tools,
- 0:25SPEAKER_00: vastly reducing help desk and customer service spend.
- 0:29But what happens when malicious actors try to use these
- 0:32co-pilots in ways they
- 0:33SPEAKER_00: weren't designed to?
- 0:34SPEAKER_00: Could a co-pilot become a data breach risk or a network vulnerability?
- 0:39SPEAKER_00: In this episode,
- 0:40SPEAKER_00: Expert Insights Head of Content Joel Witts speaks to Ben Kleiger,
- 0:44SPEAKER_00: CEO and co-founder of Zenity,
- 0:46SPEAKER_00: an application security platform for enterprise co-pilots.
- 0:51It'd be great to start just with an introduction to yourself
- 0:54and your kind of
- 0:54SPEAKER_01: background and how you got started with Zenity.
- 0:58SPEAKER_02: Michael, my co-founder.
- 0:59SPEAKER_02: We both started Zenity four years ago.
- 1:01We've been working in the cyber security space for our entire
- 1:05professional career,
- 1:06SPEAKER_02: so over 20 years right now,
- 1:07SPEAKER_02: which is insane to think about a number,
- 1:09SPEAKER_02: but that's true.
- 1:12SPEAKER_02: Working in the security space for many,
- 1:13SPEAKER_02: many years,
- 1:14SPEAKER_02: started way,
- 1:15SPEAKER_02: way back,
- 1:15SPEAKER_02: and many years of product management prior to starting the company.
- 1:19SPEAKER_02: I was with Microsoft and so was Michael.
- 1:21SPEAKER_02: I was leading their cloud security product group,
- 1:23SPEAKER_02: and it's a product that we incubated internally.
- 1:26SPEAKER_02: called it Azure Security Center.
- 1:28SPEAKER_02: It's now rebranded as Microsoft Defender for Cloud.
- 1:31SPEAKER_02: Michael was one of my closest friends and peers back then.
- 1:34SPEAKER_02: He was in the security CTO team.
- 1:35SPEAKER_02: He was in charge of innovation,
- 1:36SPEAKER_02: trying to find cool attack surfaces,
- 1:39SPEAKER_02: new problems that Microsoft should address.
- 1:41SPEAKER_02: As a vendor, he spent his time around the IoT security and API security.
- 1:45SPEAKER_02: Most predominantly before we left,
- 1:47Michael was helping Microsoft to think about what it means to
- 1:51secure cities and
- 1:52SPEAKER_02: development.
- 1:53SPEAKER_02: at scale, right?
- 1:54First,
- 1:54of course,
- 1:55internally,
- 1:55but also working with some of Microsoft's largest accounts.
- 1:58SPEAKER_02: When we went to our own entrepreneurial journey and we started to think,
- 2:01SPEAKER_02: okay,
- 2:02SPEAKER_02: so what are we going to do?
- 2:03We were really inspired by that and inspired by this whole
- 2:06motion of anyone can build.
- 2:09SPEAKER_02: Everyone is a developer.
- 2:10It started like five or six years ago with the citizen
- 2:13development and low-code,
- 2:14no-code motion.
- 2:16SPEAKER_02: And we decided that spending our time
- 2:19around bringing security principles and governance to
- 2:22non-professional development.
- 2:23SPEAKER_02: That's what we want to do.
- 2:25SPEAKER_02: So that's how we started Xanity four years ago.
- 2:26SPEAKER_02: A lot has happened, of course, in the world, in the company sense.
- 2:30SPEAKER_02: We are today over 100 people spent across Israel and the US.
- 2:34SPEAKER_02: We raised our Series B just six months ago.
- 2:37SPEAKER_02: And of course, generative AI has taken a very major
- 2:41SPEAKER_02: a place in our journey as well.
- 2:42SPEAKER_01: Yeah.
- 2:43SPEAKER_02: Because when you look at enterprises and how they adopt generative AI,
- 2:46SPEAKER_02: it relies heavily on low-code,
- 2:48SPEAKER_02: no-code,
- 2:48SPEAKER_02: and it also pushes the same idea that anyone can.
- 2:53SPEAKER_02: Anyone can create an automation,
- 2:55SPEAKER_02: anyone can create an application,
- 2:56SPEAKER_02: and anyone can create today an AI agent.
- 2:59SPEAKER_02: How do you do it with low-code, no-code?
- 3:00SPEAKER_02: So that was a natural evolution for us.
- 3:03SPEAKER_02: And yeah, we're very proud at the current place.
- 3:06SPEAKER_02: It's a very exciting week.
- 3:07SPEAKER_01: Yeah.
- 3:07SPEAKER_01: So you started four years ago,
- 3:09SPEAKER_01: you know,
- 3:09SPEAKER_01: and now this year at RSA,
- 3:11SPEAKER_01: everyone's talking about agentic AI,
- 3:13SPEAKER_01: you know,
- 3:14so how have you sort of seen the industry go from kind of not
- 3:17talking about this at
- 3:18all to it's all anyone is talking about and kind of what's that
- 3:21meant for you guys
- 3:22SPEAKER_01: at Vanity?
- 3:22SPEAKER_02: So obviously, it's very meaningful for us, right?
- 3:26SPEAKER_02: Of course.
- 3:26SPEAKER_02: And we are starting four years ago,
- 3:28thinking about this trend of anyone can and what does it mean
- 3:31for security.
- 3:32SPEAKER_02: We always like to say that we sort of saw where the puck is heading.
- 3:36SPEAKER_02: Because, yeah, of course, we started early.
- 3:38SPEAKER_02: No one called it agent EKI.
- 3:39SPEAKER_02: No one thought about AI agents back then.
- 3:42SPEAKER_02: chatbots, and people that build automations.
- 3:45SPEAKER_02: With the introduction of Generative AI,
- 3:47SPEAKER_02: it just made our journey much more exciting and much more,
- 3:52SPEAKER_02: let's call it crazy as well,
- 3:54SPEAKER_02: because it's changing the world.
- 3:57Generative agents are changing the world,
- 3:59and we're very lucky to be in this position.
- 4:01SPEAKER_02: We're happy about it.
- 4:02SPEAKER_02: Of course,
- 4:03SPEAKER_02: it's good for the business,
- 4:04and we're excited to play such a meaningful role in the
- 4:07enablement of this journey
- 4:09SPEAKER_02: because as my customers tell me,
- 4:11SPEAKER_02: Like, this is the new Excel.
- 4:13SPEAKER_02: No, who is not using Excel, right?
- 4:14Every kid knows how to create a spreadsheet and calculate
- 4:17formulas and whatnot.
- 4:19SPEAKER_02: And agentic AI is the same,
- 4:21SPEAKER_02: and it's only going to get much more and more and more popular,
- 4:24SPEAKER_02: right?
- 4:24We do believe that our role in the journey of making sure that
- 4:28people adopt these
- 4:29SPEAKER_02: technologies and that people build AI agents securely,
- 4:33SPEAKER_02: it's fundamental to the success of the underlying trend,
- 4:36SPEAKER_02: right?
- 4:36SPEAKER_02: So to sum up everything, very exciting timing.
- 4:40So what are some of the risks and some of the challenges that
- 4:42you're helping
- 4:43SPEAKER_01: organizations to solve?
- 4:44SPEAKER_01: You know, I guess what's the security risk with AI agents?
- 4:48SPEAKER_01: Of course.
- 4:48SPEAKER_02: So we first need to remember that AI agents at their core,
- 4:52they're only becoming powerful as much as you customize them
- 4:56and as much as you're
- 4:57SPEAKER_02: connecting them to tools and to more knowledge bases.
- 5:01SPEAKER_02: And also, of course, they are autonomous, right?
- 5:03SPEAKER_02: As they're becoming more and more autonomous.
- 5:05SPEAKER_02: As an attacker, think about the number of
- 5:08SPEAKER_02: employees,
- 5:09SPEAKER_02: so sort of speak,
- 5:09SPEAKER_02: to say that I can go after,
- 5:11SPEAKER_02: go after your user,
- 5:12SPEAKER_02: or after your credentials,
- 5:13SPEAKER_02: or after your files,
- 5:15SPEAKER_02: whatnot.
- 5:16SPEAKER_02: I can go after your employees.
- 5:17SPEAKER_02: And your employees are the agents that you interact with, that you build.
- 5:21SPEAKER_02: So it's a crazy opportunity for an attacker.
- 5:25SPEAKER_02: From a threat vector perspective,
- 5:28it's about detecting and preventing prompt injections by being
- 5:32able to see if
- 5:33SPEAKER_02: knowledge bases are poisoned in any way,
- 5:36SPEAKER_02: or if attackers are using both agentic,
- 5:38SPEAKER_02: but also targeting agentic platforms for remote code execution.
- 5:42SPEAKER_02: So it's all about
- 5:44looking beyond just the interface level of how people are
- 5:48interacting with a
- 5:50SPEAKER_02: chat-like agent and actually look deeper into,
- 5:52SPEAKER_02: well,
- 5:53SPEAKER_02: what is this agent doing and why is it doing what it's doing?
- 5:56Is it doing that because it was built in a way that telling him
- 5:59to do that and that
- 6:00SPEAKER_02: involves us doing vulnerability scanning in build time?
- 6:03SPEAKER_02: Are there
- 6:03like exploits that people can take control of but also like um
- 6:09is the agent
- 6:10starting to to behave or interact in a way that it's not
- 6:12supposed to do because
- 6:13someone made him do that because of all of the things that i
- 6:16mentioned prompt
- 6:17injection and rock poisoning and and and whatnot right so we
- 6:20take an in-depth
- 6:20approach of actually profiling the activities that agents are
- 6:24doing all of their
- 6:24SPEAKER_02: agent to agent communication
- 6:27SPEAKER_02: manipulations that they're doing on systems, files, data in general.
- 6:31SPEAKER_02: So it's an agent, very deep inspection that we're doing.
- 6:34SPEAKER_01: Because I guess the challenge is figuring out what's safe and what's not,
- 6:38SPEAKER_01: or what's malicious and what isn't with the AI agents,
- 6:40SPEAKER_01: and they don't always behave as you expect,
- 6:41SPEAKER_01: I suppose.
- 6:42SPEAKER_02: Correct.
- 6:43And because we also combine the build time visibility and we
- 6:46actually see where the
- 6:47SPEAKER_02: agent,
- 6:47SPEAKER_02: how is it built?
- 6:49SPEAKER_02: What is it supposed to do at any given point?
- 6:51SPEAKER_02: When we sort of,
- 6:52SPEAKER_02: let's say,
- 6:53SPEAKER_02: compare it with what it's actually doing in runtime,
- 6:56SPEAKER_02: then all of a sudden you have much more context.
- 6:57SPEAKER_02: You're like,
- 6:58SPEAKER_02: okay,
- 6:58SPEAKER_02: it started to use not Ben's identity,
- 7:01SPEAKER_02: but Michael's identity,
- 7:02SPEAKER_02: or it started to communicate with an agent.
- 7:04SPEAKER_02: that is not part of its configuration,
- 7:06SPEAKER_02: or people didn't build it in a way that it's supposed to do that,
- 7:09SPEAKER_02: why?
- 7:10SPEAKER_02: Why did that happen?
- 7:11And then you can start and go deeper and investigate where
- 7:13that's because it is a
- 7:15SPEAKER_02: prompt injection attack that we found,
- 7:17SPEAKER_02: or is it because that Ben gave his user to his,
- 7:20SPEAKER_02: I don't know,
- 7:21SPEAKER_02: whatever colleague,
- 7:21SPEAKER_02: and that colleague changed the configuration of the agent,
- 7:24SPEAKER_02: which is maybe legitimate,
- 7:25SPEAKER_02: maybe it's not,
- 7:26SPEAKER_02: right?
- 7:27SPEAKER_02: That's what we have.
- 7:30You're listening to the Expert Insights Podcast,
- 7:33home of leading cybersecurity conversations.
- 7:36Subscribe now to get insights from top cybersecurity leaders
- 7:39straight to your inbox.
- 7:41And do you have like a sort of a good example or good case
- 7:44study of how the
- 7:45SPEAKER_01: platforms worked in practice to kind of help an organisation?
- 7:48SPEAKER_02: I think last year at Black Hat,
- 7:50Michael gave a talk about showcasing how easy it is to bypass
- 7:56one of the largest
- 7:57SPEAKER_02: ecosystems,
- 7:58which is the Microsoft 365 Copilot with a single email that
- 8:02we're sending to you
- 8:04SPEAKER_02: that it can easily bypass all of the security mechanisms,
- 8:07SPEAKER_02: email security,
- 8:08SPEAKER_02: data classification,
- 8:09SPEAKER_02: whatnot.
- 8:10SPEAKER_02: And by the email getting through the data within the email,
- 8:13SPEAKER_02: whether it's encrypted,
- 8:15SPEAKER_02: hidden,
- 8:15SPEAKER_02: whatnot,
- 8:16SPEAKER_02: it gets into
- 8:17your RAG architecture,
- 8:19enabling us to completely bypass the instruction of Copilot.
- 8:23SPEAKER_02: That was just one example that we showed at Black Hat.
- 8:26SPEAKER_02: We're seeing very similar things across.
- 8:27SPEAKER_02: It's not a Microsoft problem, just to be clear.
- 8:30SPEAKER_02: It's a problem of agentic platforms.
- 8:32SPEAKER_02: and the way that they're implemented.
- 8:34SPEAKER_02: We're seeing that across vendors, across industries.
- 8:38SPEAKER_02: These are very real things that we're already seeing in the wild.
- 8:41SPEAKER_02: So the attack vectors can be an email,
- 8:44SPEAKER_02: it could be a Teams invite,
- 8:45SPEAKER_02: a Slack message,
- 8:46SPEAKER_02: a calendar invite that people send your way,
- 8:48SPEAKER_02: whether it's in Gmail or in Outlook,
- 8:50SPEAKER_02: it doesn't really matter.
- 8:52SPEAKER_02: Once that is in and that it's within the organization's boundary,
- 8:56SPEAKER_02: as an attacker,
- 8:57SPEAKER_02: I mean,
- 8:57SPEAKER_02: I just need your user
- 8:59SPEAKER_02: to ask a pilot to summarize their emails or to ask Gemini to go and fetch
- 9:04information about what was the latest transaction that we did
- 9:07with a certain
- 9:08contractor and we are taking a control over your assistant
- 9:12right over your agentic
- 9:13platform so in terms of the use cases is it in organizations
- 9:16using ai internally is
- 9:17it customer facing ai platforms like ticket booking systems or
- 9:21something like that
- 9:22where you're asking the chatbot to book a ticket you know or
- 9:24are you covering the
- 9:25SPEAKER_01: whole
- 9:26SPEAKER_01: It's both.
- 9:27SPEAKER_02: It's both.
- 9:27Like we cover external facing agentic platforms or agents as
- 9:31well as internal,
- 9:31SPEAKER_02: because the risks do include insider risk as well as external attackers.
- 9:36SPEAKER_02: So it's both.
- 9:38I will say that we're seeing more and more usage and
- 9:41development of externally facing AI agents.
- 9:44SPEAKER_02: And of course,
- 9:45these are very appealing for an attacker as a way in that is
- 9:48not monitored properly
- 9:50SPEAKER_02: if you don't have Xenity,
- 9:51SPEAKER_02: right?
- 9:52SPEAKER_01: And it's interesting, isn't it?
- 9:53Because I suppose it's different from traditional security in
- 9:55that you're sort of
- 9:56SPEAKER_01: thinking of,
- 9:56SPEAKER_01: with prompt engineering,
- 9:57SPEAKER_01: I mean,
- 9:58SPEAKER_01: how can you trick this thing into giving it what I want?
- 10:00SPEAKER_01: How can you pressure it?
- 10:01SPEAKER_01: Give me this code or, you know.
- 10:04SPEAKER_02: I think, I agree.
- 10:05SPEAKER_02: I think that the problem with that some folks in the industry still are
- 10:09not fully getting is that you can't really fight off prompt
- 10:15injections because you
- 10:16cannot predict any possible prompt that a very creative
- 10:19attacker can come up with.
- 10:21SPEAKER_02: If we try to like compare it to previous worlds,
- 10:24SPEAKER_02: like we originally tried to fight malware with antivirus and just,
- 10:28SPEAKER_02: you know,
- 10:28SPEAKER_02: blacklisting,
- 10:29SPEAKER_02: you know,
- 10:30SPEAKER_02: file hashes,
- 10:31SPEAKER_02: that failed.
- 10:32And obviously there was a new approach endpoint protection in
- 10:35India that has taken
- 10:36SPEAKER_02: a behavioral approach to detect malware, right?
- 10:39SPEAKER_02: Or bad, let's call it, a software that gets into your organization.
- 10:43SPEAKER_02: Injection, you can't predict human creativity.
- 10:45SPEAKER_02: You just can't.
- 10:46SPEAKER_02: And you can always find a way to trick the agent.
- 10:50SPEAKER_02: And that's why you need a different approach,
- 10:52SPEAKER_02: which is not just looking,
- 10:52SPEAKER_02: hey,
- 10:53SPEAKER_02: I'm going to be at the gate,
- 10:54SPEAKER_02: and every prompt,
- 10:54SPEAKER_02: I'm going to inspect it.
- 10:56SPEAKER_02: I'm going to see if it's a pound injection attack or not.
- 10:59SPEAKER_02: No,
- 10:59you need to go deeper than that and understand the full context
- 11:02and what the agent
- 11:02is doing and how it's communicating in order to really find
- 11:05many more types of
- 11:06SPEAKER_02: attacks.
- 11:07SPEAKER_02: It will never be 100 bulletproof, never.
- 11:09SPEAKER_02: And we also need to understand that as an industry,
- 11:11SPEAKER_02: it's about,
- 11:12SPEAKER_02: as always with security,
- 11:13SPEAKER_02: it's about layers.
- 11:14It's about making sure that you can detect it fast,
- 11:16that you can react to it even faster,
- 11:18right?
- 11:19SPEAKER_02: And that's where we put our focus,
- 11:20especially when we're working with some of the largest
- 11:23enterprises in the world.
- 11:24SPEAKER_02: So we understand that.
- 11:25SPEAKER_01: Are you working with the AI developers as well,
- 11:28SPEAKER_01: like Microsoft and their team developing Copilot?
- 11:32We're always working whenever we find a vulnerability or
- 11:35whenever we find an
- 11:36SPEAKER_02: opportunity to improve the underlying platform.
- 11:39SPEAKER_02: Yes, of course, we disclose it.
- 11:40SPEAKER_02: We follow all of the responsible disclosure guidelines.
- 11:44SPEAKER_02: Yes, it's part of our responsibility.
- 11:46But there's always something new and attackers are just getting
- 11:49more and more.
- 11:50SPEAKER_02: sophisticated.
- 11:51SPEAKER_02: It's not just because those platforms are, again, insecure, right?
- 11:54No,
- 11:55they're actually doing a really good job,
- 11:57like Microsoft and Google and OpenAI.
- 12:00SPEAKER_02: Kudos to them, really.
- 12:01They're doing an amazing job to securing their platforms,
- 12:03but again,
- 12:04it's not bulletproof.
- 12:05And organizations need to take their part in this yet another
- 12:08shared responsibility
- 12:09SPEAKER_02: model that it's on them to make sure that their perimeter is properly
- 12:14SPEAKER_02: monitored,
- 12:15that they have visibility to what people are using and how
- 12:17they're using it to make
- 12:18SPEAKER_02: sure that they can control the risk and of course reduce it over time.
- 12:22SPEAKER_01: I suppose,
- 12:23SPEAKER_01: circling back to what we were chatting about at the start,
- 12:25sort of flip side of the coin with the amount of noise in the
- 12:28space of Identic AI
- 12:29and how that's great is that there's also a lot of other
- 12:31vendors popping up saying
- 12:33SPEAKER_01: they do the same thing.
- 12:34Do you sort of see yourself as having competition or are you
- 12:36really the only ones
- 12:37SPEAKER_01: doing this well in terms of security?
- 12:39SPEAKER_02: So it's a very good question.
- 12:40SPEAKER_02: I think it's one of those moments,
- 12:43SPEAKER_02: like I've been doing this for a while,
- 12:45SPEAKER_02: that a lot of people are really trying to chime in.
- 12:48SPEAKER_02: And really,
- 12:48SPEAKER_02: like,
- 12:49SPEAKER_02: and of course,
- 12:49SPEAKER_02: it's because of the business opportunity,
- 12:50SPEAKER_02: but also because of that,
- 12:52SPEAKER_02: the fact that...
- 12:52SPEAKER_01: A couple of years ago,
- 12:54SPEAKER_01: zero trust,
- 12:54SPEAKER_01: right?
- 12:54SPEAKER_01: Everyone suddenly had a zero trust path.
- 12:56SPEAKER_02: Yeah, yes, but it's also because of the fact that...
- 13:00SPEAKER_02: This is happening so fast.
- 13:02SPEAKER_02: Just think about ChatGPT, everyone are using it today.
- 13:05SPEAKER_02: It was just introduced to the world.
- 13:07SPEAKER_02: How long ago?
- 13:08SPEAKER_02: Two years ago.
- 13:09SPEAKER_02: So it's still a very new technology from one end,
- 13:12SPEAKER_02: but on the other hand,
- 13:13SPEAKER_02: so many organizations and individuals are already adopting it.
- 13:16SPEAKER_02: So the security community and vendors, of course,
- 13:19SPEAKER_02: coming together to making sure that people do that securely.
- 13:22SPEAKER_02: But yes,
- 13:22SPEAKER_02: of course,
- 13:23to also take the opportunity of the business opportunity that
- 13:26comes with it,
- 13:26SPEAKER_02: right?
- 13:27SPEAKER_02: Many of the vendors today,
- 13:28like there was a first wave of companies that was focused on
- 13:31securing LLMs,
- 13:32SPEAKER_02: right?
- 13:33SPEAKER_02: And customer LLMs that organizations are...
- 13:35SPEAKER_02: are building, training their models.
- 13:37SPEAKER_02: Like, how do you make sure that people cannot compromise your model?
- 13:40SPEAKER_02: And there are amazing companies that are doing that.
- 13:42I think that agentic AI is changing the landscape because LLMs
- 13:46are becoming more
- 13:48SPEAKER_02: and more commoditized.
- 13:49SPEAKER_02: Do organizations need to continue building their own LLMs?
- 13:53SPEAKER_02: I don't know.
- 13:54SPEAKER_02: Like, we'll see, right?
- 13:55But will organizations and will people continue to build and
- 13:58adopt agents?
- 13:59SPEAKER_02: 100%.
- 14:00SPEAKER_02: Yeah.
- 14:01SPEAKER_02: Yes.
- 14:01SPEAKER_02: Again, it's the new Excel, right?
- 14:03SPEAKER_02: So yes, for sure.
- 14:04And there,
- 14:05there are also all sorts of vendors that are chiming in,
- 14:06are trying to solve this.
- 14:08Some vendors are trying to solve this from,
- 14:09oh,
- 14:09this is an identity problem.
- 14:11SPEAKER_02: Some vendors are saying this is a data security problem.
- 14:14SPEAKER_02: We believe that you need to actually flip it around.
- 14:16SPEAKER_02: You need to focus on the entity,
- 14:18SPEAKER_02: which is the agent,
- 14:19SPEAKER_02: and then you can examine,
- 14:20SPEAKER_02: okay,
- 14:20SPEAKER_02: so what is the data problem that comes with it?
- 14:22SPEAKER_02: And what is the identity problem that comes with it?
- 14:25SPEAKER_02: And how you can solve it
- 14:26per agent and per profile or purpose that that agent is being
- 14:32built for and used for.
- 14:33SPEAKER_02: So I think that, I hope that we are unique in that sense.
- 14:40SPEAKER_02: But again, very exciting timing.
- 14:42SPEAKER_01: Yeah, absolutely.
- 14:43SPEAKER_01: How do you sort of see the market changing over the next 12 months?
- 14:46I think that the world is going through some changes right now
- 14:52across the world.
- 14:53One of the things that I think are more stable are the fact
- 14:57that AI agents are not
- 15:00SPEAKER_02: going anywhere.
- 15:01SPEAKER_02: they're only going to be adopted even more, right?
- 15:04SPEAKER_02: With a lot of,
- 15:05SPEAKER_02: you know,
- 15:05SPEAKER_02: let's call it macro-economy changes going on in the world,
- 15:08SPEAKER_02: companies will look to be more efficient,
- 15:11SPEAKER_02: to have a competitive edge.
- 15:13SPEAKER_02: The way to do it is with adopting AI agents, right?
- 15:15So I actually think it's one of the things that are going to be
- 15:18more,
- 15:19let's call it...
- 15:22SPEAKER_02: Yeah, stable again, right?
- 15:23SPEAKER_02: In our world and especially in the tech scene.
- 15:25SPEAKER_02: And with that, the adoption is only going to grow, right?
- 15:29Security plays a very critical part of it to making sure that
- 15:33not only that
- 15:33SPEAKER_02: companies actually feel safe to do it,
- 15:35SPEAKER_02: because it's again,
- 15:36SPEAKER_02: it's a new technology.
- 15:37SPEAKER_02: It's just like they're adopting it while pushing it,
- 15:40SPEAKER_02: while trying to understand what it means to secure it,
- 15:42SPEAKER_02: right?
- 15:42SPEAKER_02: So I think security plays
- 15:44SPEAKER_02: a very significant role security in governance plays,
- 15:46SPEAKER_02: a very significant role in the enablement of this technology.
- 15:51And I'm very bullish on Xenity and on the agency security space
- 15:56in general.
- 15:57What sort of excites you most about the journey ahead for you
- 16:00guys,
- 16:01the broader security space?
- 16:02SPEAKER_02: Honestly, the impact.
- 16:05SPEAKER_02: Yes, we're not a pro bono thing.
- 16:09SPEAKER_02: We're building a company.
- 16:10We're building what both Michael and I believe and the board
- 16:13believes is the next
- 16:15big and important company in cybersecurity because it has room
- 16:18for its own space,
- 16:20SPEAKER_02: its own expertise.
- 16:20SPEAKER_01: It's a new category.
- 16:21SPEAKER_02: It's a new category for sure.
- 16:24SPEAKER_01: Who do you look at in the space that you aspire to be?
- 16:26SPEAKER_01: Are you looking at the way CrowdStrike's done with EDR?
- 16:29I'm a huge admirer of Wiz and of CrowdStrike and of the
- 16:34SentinelOne people.
- 16:34SPEAKER_02: I think they all built excellent companies.
- 16:37SPEAKER_02: I know some of those people well, some of them are friends.
- 16:41SPEAKER_02: Yeah.
- 16:42SPEAKER_02: ZScale,
- 16:43SPEAKER_02: amazing companies,
- 16:44SPEAKER_02: really,
- 16:44that all started from thinking about where the future is going
- 16:48or where companies
- 16:49SPEAKER_02: are going in the future,
- 16:50SPEAKER_02: I'm sorry.
- 16:51SPEAKER_02: It's always about that when you're trying to create a category.
- 16:53SPEAKER_02: It's not about, I'm going to solve that new feature, right?
- 16:57SPEAKER_02: Or build or only offer that specific pain that someone has, right?
- 17:01SPEAKER_02: Yeah.
- 17:01SPEAKER_02: I think that AI agents are
- 17:03SPEAKER_02: like that, right?
- 17:04SPEAKER_02: I mean, there's no question that they're changing the world.
- 17:06SPEAKER_02: There's no question that everyone is going to use it.
- 17:08SPEAKER_02: There's no question that everyone is going to build.
- 17:10SPEAKER_02: Agents as well.
- 17:11SPEAKER_02: We have employees internally at Xena team included.
- 17:13SPEAKER_02: Like we have agents running, you know, that we develop, right?
- 17:15SPEAKER_02: It's changing the industry.
- 17:17So I'm super,
- 17:18super excited about building a new category here,
- 17:21which is,
- 17:22it's not easy.
- 17:23SPEAKER_02: But it's also what keeps us so motivated to do this.
- 17:31SPEAKER_02: And also the big impact that it makes, right?
- 17:32SPEAKER_02: Because it is really going to be adopted by everyone.
- 17:35SPEAKER_02: It's going to change the way that we work,
- 17:38SPEAKER_02: whether we're talking about legal firms,
- 17:41SPEAKER_02: insurance companies,
- 17:42SPEAKER_02: bankers.
- 17:43SPEAKER_02: Everyone is going to use it.
- 17:44So the impact,
- 17:46I think,
- 17:46is unique,
- 17:47even compared to previous security categories.
- 17:51SPEAKER_01: Yeah.
- 17:52SPEAKER_01: Fantastic, thank you so much.
- 17:53SPEAKER_02: Thank you, I appreciate it.
- 18:06This podcast was produced by Expert Insights,
- 18:09home of leading cybersecurity conversations.
- 18:12Subscribe now to get insights from top cybersecurity leaders
- 18:15straight into your inbox.
- 18:17Why not check out our weekly decrypted cybersecurity briefings
- 18:20and our Game
- 18:21SPEAKER_00: Changers conversations with innovative individuals.