Ep41 | Data Harbor Platform Launch: Why it Matters
November 17, 2026
Recorded the day Virtru launched the Data Harbor Platform, Matt Howard and Cole Grolmus hash out why AI agents move data security from the back of the line to the front of the room, and why the answer is "and," not "or."
Every organization is under pressure to adopt AI agents faster. Every security team is under pressure to make sure that doesn't go wrong. Both pressures land on the same thing: the data.
Virtru CMO Matt Howard sat down with Cole Grolmus, founder and CEO of Strategy of Security, minutes after the Data Harbor Platform launch. No script, no slides. Two people who have spent their careers in security, talking about where the industry goes next.
They get into why data security has been "important but never urgent" for twenty years, why agents end that, and what happens to a pilot program the moment it touches the data that actually matters. Cole is a student of strategy, so he also names the market dynamic most vendors don't want to say out loud.
Read transcript Hide transcript
Matt Howard
Welcome. My name is Matt Howard. I'm the chief marketing officer at Virtru, and, I have the pleasure of being joined here today by Cole Gromis, the founder and CEO of Strategy on Security.
Cole, exciting day for Virtru. We just wrapped up the Data Harbor product launch and was kinda just hoping to get a quick sort of readout from you. What did you see? What did you think was interesting? And what's exciting?
Cole Grolmus
Look. I mean, I think this is a super exciting day, like, long time coming. I mean, that that...
And I mean that in the best way possible. Like, I have this, I don't know, special affinity for some reason of companies who, you know, are not kind of fly by night. Let's pour as much venture capital as we can into this, you know, maybe take an idea that's been out there before and, like, speed run the whole thing. This this has been many many stages in the making and, like, a ground up, you know, as I read on the wall out there, like, brick by brick type of a build.
To, you know, to hit a point like we did today is is super important. So I, you know, I I I wanna first, like, foundationally start with that. Like, that's a that's a really big deal, and it's a really big accomplishment.
Matt Howard
Yeah. Thank you very much. I mean, I mean, as as some people know, Virtru is a fifteen year old company that's been working for, you know, that long of a period to really do something that's, you know, genuinely different and innovative in the world of security, and that's creating sort of data that's capable of protecting itself with policy that's cryptographically bound to the object. We've proven that out over a long period of time with email and file workflows. But today, launching the Data Harbor platform and sort of exposing those capabilities to third party developers with APIs and SDKs and sort of showing the world what's possible when we enable others to inherit the power of TDF.
It feels really good. And to your point, it's been a long time coming, but we're here at this moment in time. And I'm just kinda curious. You see a ton of real estate in cyber these days, and you know a ton of vendors, you know, from the identity guys to the network guys to the DSPM guys. And, you know, there's just so much going on in in the cyber game these days.
Where do you think the sort of data security fits in that sort of worldview? How important is it? What's the... You know, what's that picture look like?
Cole Grolmus
Yeah. I mean, I think it's super important in a way that it never never has been or hasn't hit this point yet. I mean, there's a reason why I'm here today. There's a reason why I'm sitting in this room is because I I very much believe in, you know, the mission, and and I'm intrigued by, you know, the specific solution here behind TDF and what Virtru has built on top of that. Like, you know, look, I I I I have I have some people who I really respect in the data security community both in this room and, you know, and and others that I work with.
And, you know, everybody's pretty self aware about that where it's like data... One of them told me, like, data security has always been something that, you know, is known to be somewhat important but never urgent, like, never a thing that we're gonna do, you know, right away. It just always, for whatever reason, seems to fall behind other aspects of security, you know, like at the endpoints, the the the security operations, identity, even, you know, things things like that. And, like, I don't think that's gonna be possible anymore. Right. I I I think I think that's where where agents are going to take us and where the the fast moving pace of the world that we're in today is going to land us. And and some of what General Nacasone said today that's very important, right, is like, I think we have to think differently about the way that we approach security, the way that we approach resilience. And I think that data security is probably the front and center of that.
Matt Howard
Yeah. It's super interesting and and, you know, grateful to kinda hear your thoughts on that.
There's no doubt that as an industry, and you know this, I've been in this game a long time, we've seen for decades now a particular game that's been played in arguably a certain way with a certain approach architecturally from identity to the endpoint, to the network, to the application. And all of that governance, all of that security, all of those controls are super important. But to your earlier point, you know, one thing people have talked a lot about data security or securing the data, you know, it's kind of been a little bit of the tail wagging the dog. And the question is, you know, how much longer can we...
Is that going to be true? And at what point does more and more of sort of the customers that are out there struggling with traditional sort of cybersecurity challenges, but yet all of a sudden faced with this entirely new reality where every one of the organizations in, you know, any industry is sort of, you know, challenged to adopt AI and agentic workflows faster for purposes of creating value for shareholders, but at the same time, fundamentally challenged with the idea of how do they govern it? How do they manage risk? Like, what is it?
Like, it's this, like, push and tug between I gotta move faster to adopt AI, and I better do it securely, and there's no alternative. All of a sudden, people are, like, asking themselves fun...
Some pretty fundamental questions. Like, how? How do I do that? And, obviously, we have a view, and and, you know, our our view that we put out on the stage today is that you need to sort of look at the data perhaps first and think about ways that you can control and govern AI by literally creating policy that's cryptographically bound to the data so the agent only gets access to the data that it's entitled to and nothing else.
Cole Grolmus
Yeah. I mean, I think I think it's like an end. Right? Like, that...
Matt Howard
That's...
Cole Grolmus
And that's the hard thing to mentally balance, and that's the nuance that we can kind of get into here in Yeah. Long form discussion. Right? Is it...
It's You got it. You said it perfectly. It's like all those other layers are important. Network security is important.
Endpoint security is important. Identity is important. Security operations underpinning all of that is important.
Your layer of the security stack. Like all of all of that matters. And so nobody nobody is sitting here saying or not many people are sitting here saying like just discard all of that. Like that's... That all has to be there still.
I think the and part...
If you peel back the questions all the way back from that, it's like, what do you control on the network? The data that flows through it. What do you control on the endpoint? That's where the data gets processed or the way that you interact with it.
Where do you control the identity? It's it's the the the person or the thing who accesses the data. Like, ultimately, all of those layers of control come back to data, but it was always like, we're gonna do those things first, and there are reasons for doing that in, a historical architectural paradigm. Like, now it gets to the end.
Like like, you have to do all those things and you have to secure the data. I think there's no way around that.
Matt Howard
Yeah. I think that's well said. And and I mean, to, the point you made, it is also important to, I think, reiterate for the sake of the conversation that we're having in the market with customers and partners and and, you know, others all the time.
It is not one or the other. It is both. I mean, absolutely, if we think about, you know, what what the control plane looks like in, like, a true sort of zero trust architecture as it would be prescribed, for example, by the Department of Defense, you gotta have all of the pieces of the puzzle properly sort of, you know, aligned from identity to endpoint network apps, and you have this sort of interesting sort of data pillar.
And this data pillar, you know, has arguably two sides to it. It's the do the discovery, do the classification, do the tagging, like make sure that you know where all of your data is, make sure that it's classified appropriately what's sensitive and what isn't sensitive. And then where we think it gets really interesting and exciting for the Data Harbor platform that we launched earlier today is the opportunity to put real controls on that data. So then when it inevitably moves downstream into any particular workflow, you have policy and security and control that that travels with the data object. Yeah. I'm kinda curious to get your thoughts on, like, the DSPM category specifically because they have... If we think about the macro cyber market, we think about data security. In the last few years, there's certainly been a lot of attention, a lot of capital raised, and some really interesting growth stories, for example, Sierra Yep.
And others.
Thoughts on DSPM as kind of an... A a rapid rising category in the data conversation and how that maybe sits in an adjacent and complementary way to, like, what we're doing at Data Harbor?
Cole Grolmus
Yeah. I mean, you know, look, like, the the the the connection and the tailwind that AI has given data security is now very clear. I mean, that was, you maybe theory a couple years ago. I don't think that's theory anymore.
I mean, that's...
You know, I I I for sure think that that's there. And you and you can see that momentum. Right? Like, there's a reason Sayra is Sayra.
You know, there's a reason that Varonis was That's right.
Was the the largest The greatest hammer.
Security action ever.
Like, incredibly Impressive company that that Yaki and the team built there.
Matt Howard
Yep.
Cole Grolmus
Right? You know, and and name many others. And like...
And I think to be clear here that, you know, this is this is the like DSPM as a category versus, you know, where Virtru sits with with Data Harbor. Like, those platforms are essential. I mean, that's one of the that is one of the the the important control layers, especially in, you know, an enterprise or, you know, a a large, public sector organization. Like, there's no question about that anymore.
Like like, you gotta have that. You gotta have a proper data security platform that goes all the way, you know, and DSPM isn't even enough either, by the way. Like, that's very important. You need to be able to find your data.
You need to classify it. You need to tag it. But, like, you know, you've you've also got to be able to do, you know, data access management around it. Like, there's there's many other layers Exactly.
Of the data security stack that are beyond just finding and tagging and doing some basic, you know, access control around it that that matter. I think the part that's missing that, you know, this is the maybe slightly more forward looking part and for sure what I think Virtru, you know, has has been early to and on top of for a long time, I think you've got to put all of those aspects of data security and identity together in a way that allows the data to protect itself. Like like that that is the secret sauce. That is the magic right there.
When when the data can protect itself, that enables a whole other layer of security that has never been possible in the time that I've been doing this.
Matt Howard
Yeah. And I'm so glad you brought that up. I mean, we we we certainly...
Gonna choose my words carefully here, but, you know, well said, and in particular, this idea that, you know, our focus has absolutely been on this idea of attribute based access control where where, theoretically, your data state is is sort of understood. A data has an attribute to sort of, you know, identify it as sensitive or not, and that attribute on that data then sort of leads downstream into a policy decision that this type of data with this type of attribute is accessible by these types of humans with these entitlements or by these types of agents with these entitlements. And the identity piece is something that we've, talked about a long time, but just very recently and not something that we've announced publicly yet, but as I think you're now well aware, we just recently announced...
Sort of closed an acquisition in exactly that space, you know, giving ourselves the opportunity to extend the story, you know, all the way from the attribute on the data where the policy is cryptographically bound to the object down to the human. I...
The human who then is basically delegating entitlements to an...
A machine, an agent, and having that sort of cryptographically sort of auditable and verifiable. So you can basically see sort of a chain of provenance all the way from the data, the way down to the machine identity with the entitlement.
That's a view of like the end to end sort of end game that we're working towards with this new acquisition, which we're going to announce later next month. But love to hear your thoughts on on sort of, you know, not just that that we're developing, but just identity in general. Like, where's the identity ecosystem today with respect to data security?
Cole Grolmus
Yeah. I mean, I think I think we've hit a point where, like, they're they're a lot more aware. Right? I mean, I'm speaking from the point of view of the identity company, and and probably the other way around.
I mean, you could make the argument too that, know, like, Varonis has been in the identity game for a long time even though it's kind of an understated Yeah. Thing in the platform. Like, it's actually, like, very strong and very impressive. Sayara made a a very large Away.
NHI acquisition, you know, and there have been several others in the in the space as well. So, I mean, kinda, you know, like, from both directions, you can see them working, towards each other, and there's, you know, I I would...
I don't even know that I would call it competitive necessarily. I mean, I think it's it's...
There's this cooperation and awareness that just, like, data and identity are joined at the hip. That's you need to be able to do that, or you need to be able to put those together.
You know, look, I I I think, like, in in the the thesis that you just put forward in the context of Virtru and the deal that you just did, just think that if you use advanced agents enough, can see where that level of detail, like down to actual machine identities or service accounts and tokens, down to attributes, specific rows in a database, a specific object in a video, for example. Like, I've seen some wild things today that you can do with TDF and the Virtru platform, and, like, humans are more forgiving. We used to... You you could sort of dream about those things before, like, in a pre agent Right. You know, world, and and it was like, oh, that'd be cool, but we don't...
Matt Howard
Why would you?
Cole Grolmus
Yeah. It's the overkill. We don't really need to necessarily, but we kinda need to now. The speed at which these things like agents traverse other layers of control, again, that still need to be there, but, like, the speed that they traverse those is very fast.
The way that they act upon data is very fast. Yeah. Like, it...
It's it's just, you know, in the way that they, like, fan out and cover multiple aspects of the of the of the data estate in one session to solve a problem is, like, true. Like, it's remarkable, but it's also scary. So I I I don't know. I mean, I sit here and I'm, like, super impressed and super terrified all all day long.
But it it...
All this is to say, like, it really it really brings the need for more sophisticated approaches to identity and data security that we have just not gotten to before. It brings them to urgency.
Matt Howard
I I think that's exactly right. And and, you know, I had an interesting conversation with a colleague of mine the other day, Adam Hajke, who runs our our solution architect team. You know, we we were just talking about sort of, know, like, you put yourself in the shoes of someone in a large enterprise today and you're faced with the current reality that they have, and it's not just sort of the rapid proliferation of zero days and the the race to sort of, like, address that reality with patching and all. That's nuts.
You also basically have this sort of pressure that we talked about. You know, the organization's got to adopt AI as quickly as possible. And make no mistake, do it responsibly, do it in a way where, you know, risk management is in full scope. So then, you know, we walk into the room, we're sort of introducing the concept of the Data Harbor platform and the capabilities that are possible with something like trusted data format, you know, to really sort of challenge somebody to think about security architecture that's fundamentally more granular, granular all the way down to the data object level.
It's like...
To your point, like, in the past, you know, people would go, it's interesting. I might wanna do that. And...
But but really, you want me to cryptographically bind policy to every object of data? That...
That's probably too much effort. Yeah. The juice isn't worth the squeeze. It sounds interesting, but I've got, you know, limited budget and higher priorities.
No. I'm good. Like, that reality isn't going to cut it anymore. Like, because at some point...
And by the way, just to be super clear, in my conversation the other day with with Adam, you know, I don't personally sit here and think for any moment, that that a sophisticated enterprise would look at their end state and say, all data needs to be protected with TDF. Yeah. It doesn't. Like, let's be real.
We're talking about data that is very, very, very highly sensitive, that is, like, ultimately the kind of stuff, you know, that creates really critical value for your shareholders. It's your intellectual property. It's it's the m and a deal. Like, you know, all that kind of stuff is ultimately so sensitive and so valuable that it does beg, the question, wouldn't you wanna have something like TDF to govern it at a very granular level so that whatever policy gets defined is enforced cryptographically wherever that object goes for as long as it lives?
Because inevitably, I might change my mind, and I wanna might revoke the access to it two weeks later after I've given it to you, for example. Yeah. Kinda curious to get your thoughts. I mean, obviously, data states are massive.
It's a small subset of the data that theoretically is really well suited for something like TDF in those high high value workflows.
Cole Grolmus
Yeah. Yeah. For sure. I mean, the...
I wanna take, a half step back on something that you said because I think it's really important context, especially in an enterprise environment. Right? Like, the time that I spent at PwC, you really learn what buying cycles look like. Right.
And, like, what...
And and forget it...
Forgetting to the commercial aspect of it. Right? Like, what does it take to make meaningful change in in an organization in the context of security? Right?
There's, like, there's, like, three things. It's enablement. It's security. You think like attack, you know, attack based, like, true cybersecurity and compliance.
Those are pretty much the only three ways that you can get anything Done. Done. A deal. Yeah.
Matt Howard
Yeah. Yeah. I mean, or even...
Cole Grolmus
Or a project, like, moving forward. Mean, even if you're talking, like, internally funding it. Right? Like, just to be able to make that happen, it's gotta be, like, at minimum one and better, you know, better two.
I don't know if I've ever seen all three. Like, it just...
It it doesn't happen that often. Right? And so, like, this is this is a really interesting time to get bigger things done like what you're talking about because, you know, what what you're saying earlier about, you know, getting...
Like doing something like this in an enterprise, like, well, you did not mention compliance. It is a reason that you would do, you know, that you would do something like like a back or something, you know, very sophisticated. But, like, I think enablement and security are just as important. Right.
Right? Like, you've got AI initiatives that are happening, and, I don't wanna be in the middle of that. Would never tell any of...
Mean, my friends are security people or compliance people. All my former clients are security people. I would never ever tell them to get in the way of an AI initiative. Not a good idea.
Matt Howard
Right.
Cole Grolmus
Bad for your career.
Matt Howard
Yeah. Career limiting. Organization limiting. It's just...
Cole Grolmus
There's there's just no there's no reason for it either. Right? And so, like, I I think short of that, what you have to do is be really creative at finding ways to work around that and align yourself with, like, the right organizational drivers to be able to get meaningful change done and do hard things that maybe had not been done before. Like, that's a way of rethinking about security.
I I...
And I...
That's why I think, like, enablement and and and the attacks that we're facing from a security standpoint are gonna be the way forward.
Matt Howard
Yeah. I mean, just to put a fine point on that, you know, real world example. I mean, we're we're in conversations with a number of large financial institutions. You know, one of them, you know, recently told us they have a bunch of, you know, very sensitive PII, and they're talking about sort of some of their early adoption of these agentic, you know, AI workflows.
And they're like, well, they're not allowed to get anywhere near the PII. Like, it's just out of scope. Like, the agents don't get near our PII. And it's like, okay.
That's that's that's a particular financial organization.
A different financial organization sort of has us in partnership with them, you know, working, you know, in a POC now where the PII is in fact the thing that the agent is accessing for purposes of automating a workflow which the bank determines is important. And they... Because it's important to the business of the bank, they want to have that value. But equally, because they understand the scope of that risk, they wanna make sure that those individual objects are properly controlled, and they wanna do it with TDF and policy that's bound to the object so that the agent is only accessing the PII that it's entitled to and not anything else.
Cole Grolmus
You just drew the line between like what like the the like shades of enablement. Right? Like like what you just described is the difference between like a mission critical, you know, truly valuable agent...
A business agent versus a toy. Right? Like, if you're treating agents like a toy, you know, you don't need PII. Right.
Or or...
I mean, it doesn't even have to be PII. Right? Like name name your favorite class...
Intensification classifier. It might be, you know, it might be intellectual property. It might be whatever it may be. Like whatever sensitive means to you.
If you want your agents to be toys and like you want your AI projects to fail, that's that's fine. Then do that. Right? But like I I I don't think that's gonna last very long.
Right? I don't I don't think that's gonna be...
Matt Howard
That's kind of like to your earlier point. That's like that organization is kind of standing in the way of AI if that's all they're doing.
Cole Grolmus
Yeah. Yeah. Yeah. Or you might be like enabling it, which is like sure, you can use, you know, you can use AI, but it's not gonna do like, the real stuff.
Matt Howard
Right.
Cole Grolmus
Like, that... That's not gonna last. It's... They're gonna have to be able to do the real stuff, and the people that get to that sooner are are gonna be better for it. And and and, like, it can. Like, there are ways to be able to do it. That's... That that is that POC. Right?
Matt Howard
Right. Yeah. Well...
So one last question, and then I'll then I'll sort of let you go. We can wrap up and go enjoy a cocktail with the rest of the, the crowd that's out there. I I am curious to get your thoughts about this idea that now is a moment in time. The, you know, the rapid proliferation of all things AI and and agents sort of creates this scenario where everybody in the industry clearly...
I mean, I think it's last count I got was like 400 plus companies in one form or another from startups to, you know, CrowdStrike and Palo and everybody in between are basically looking at this idea of how do you govern, you know, agents and their behavior. And so much of what I'm seeing, and I'm just curious if you're seeing the same thing, seems to be focused on this idea that you want to govern and control the agent sort of in the runtime, on the endpoint, at the application abstraction with some type of a harness. Makes perfect sense.
And and, you know, if if I were a a CISO responsible, I'd do the same thing.
And all we're sort of doing today with the Data Harbor launch is wanting the world to sort of really step back and sort of see that in addition to doing that, there also is another opportunity to get really good governance in place by putting policy that's cryptographically bound to objects so that you are basically achieving the same end game of governing the agent behavior from the data's perspective as opposed to doing it all on the endpoint in the runtime. I'm just curious. Is that reasonable? Do you see the difference? Kind of is it healthy? Like, what's your thoughts there?
Cole Grolmus
Yeah. You know, look. I'm I'm a student of strategy, and I do security. So hence, know, hence the strategy of security part. Right? And, like, one thing that I love is blue oceans. Right.
Matt Howard
Right? You and me both. It's it's hard to find those, and it's it's beautiful when you do. Right?
Cole Grolmus
What what you are describing is is the red ocean. Right? And, like, that's fine. Like, all...
Matt Howard
You know, run time We're a 100 companies.
Cole Grolmus
Runtime security and all the other aspects of things around AI security like like AI identity. I mean, you can kinda go through your, you know, your your AI... Your emerging AI security stack. Like, all that needs to exist. That's fine. And, like, I I... You know, there's a lot of awesome companies out of there. Some of them are my clients. Like, there's there there there there's there's good stuff happening, but, like, it's brutal out there.
If you were to sort of take the adoption curve, that's what it's like to be in a nascent emerging market. It's like an inherent property of it is you're going to get a lot of competition.
You know, like market entry is going to be really easy. They're just going to bang heads for a while, and that's going be how it is.
Matt Howard
Definition red ocean.
Cole Grolmus
Yeah. Yeah. That fully. Right? And so, you know, this is a blue ocean. That's like the beautiful thing that we're sitting on here is and practically speaking, like even from putting my security hat back on, right, like that's what I was trying to get at is this this side of the control plane needs to exist too. Right.
Like again, not an and, you know, not not an or, it's an and.
Matt Howard
That's right.
Cole Grolmus
All that stuff needs to happen. That's right. All all of the layers of technology that we've used for our entire career across the entire stack, that all still needs to be there and adapt to to agents too.
I think we need this piece Yep.
Around data security that protects... Allows the data to protect itself. I think that's really important.
Matt Howard
Yeah. I mean, I I think that's well said. I appreciate that perspective. And and certainly, you know, our view is we are not naive.
We certainly think there's a ton of really powerful innovation going on out there where really great companies are delivering credible solutions to the market to kind of govern agents in the runtime on the endpoint and more power to them and the value that that offers. At the same time, we just think there's a complementary capability of putting policy sort of on the data to to to your point, not not instead of, but in addition to. And that way, it's not one or the other, but both. That gives you kind of a control plane from end to end to kind of really rest comfortably at night knowing that your agents are doing what they need to be doing in order to help create value for shareholders, but at the same time knowing that, you know, your risk is effectively managed and unintentional things aren't going to happen.
But super grateful for your time, for your expertise, and for the partnership. Thank you for being here today at the Data Harbor platform launch. We will be sharing this on the Virtru website and invite any of the folks that might be watching this to absolutely take a look at the the Virtru website to see all of the latest and greatest with respect to the Data Harbor platform launch that we just completed today. So, Cole, thanks for being here.
Cole Grolmus
Thanks for having me.
Related Resources
Get expert insights on how to address your data protection challenges
Ep40 | Virtru + DataXRay - From Discovery to Defense: Governing Sensitive Files at Scale
Ep39 | Virtru + Rocket.Chat: The Last Mile of Zero Trust - Governing the Collaboration Layer
Ep38 | The Lean CISO in the Age of Agentic AI
Ep37 | The Core of Zero Trust: Protecting Data While Powering the Mission
Ep36 | ACP 240 in Action: Data-Centric Security for the Mission Partner Environment
Ep35 | Achieving CMMC Level 2: Insights from the OSC, Consultant & Assessor
Ep34 | Navigating CMMC: Architecture, Options, and Real Security for the DIB
Ep33 | Digital Fortifications: How Taiwan and Ukraine are Advancing Tech Security Under Threat
Ep32 | Data-Centric Zero Trust for Missions & Markets
Ep31 | Why Federal Cyber Leaders are Coming to Virtru
Ep30 | Beyond the SEG: Building Superior Email Security with Abnormal & Virtru
Ep29 | Crypto Under Siege: The Thin Line Between Digital and Physical Security
Book a Demo
Become a Partner
Contact us to learn more about our partnership opportunities.
Become a Compliance Champion
Contact us to learn more about our partnership opportunities.