Barracuda to MSPs on AI security: Don’t wait for customers to ask

**Robert Dutt:** Hello, and welcome to In The Channel from ChannelBuzz.ca, bringing news and information to the Canadian IT channel for the last sixteen years. I’m Robert Dutt, editor at ChannelBuzz.ca and your host for the show.

AI security is quickly becoming another challenge for MSPs to help their customers navigate. But there’s an interesting wrinkle. Many SMB customers may not even know how much AI is already being used inside their organizations, what information employees are putting into those tools, and what risks that creates.

Welcome to the fun world of Shadow AI.

Barracuda believes that gives MSPs an opportunity to get ahead of the conversation.

The company recently introduced new AI security capabilities aimed at discovering Shadow AI, monitoring how AI tools are being used, and helping organizations prevent sensitive information from finding its way into large language models.

I recently caught up with Brian Downey from Barracuda to talk about the new offerings, some new research into the AI skills gap among mid-market businesses, and why he sees parallels between today’s AI governance opportunity and the early days of managed security.

We also discuss how MSPs can turn visibility into an advisory conversation, why they shouldn’t wait for customers to start asking about AI governance, and whether securing AI could ultimately become as standard a part of the managed services stack as backup or MDR.

Let’s get right into it, my chat with Brian Downey.

**Robert Dutt:** Brian, thanks for taking the time. I appreciate it.

**Brian Downey:** No, I appreciate it, Rob. It’s great to be here.

**Robert Dutt:** You were the voice of the Barracuda One announcement back in March, and this feels like kind of the second chapter of that story. Can you maybe start by setting the scene a bit? What is AI Data Security from Barracuda? And can you talk about why you felt this is kind of a standalone product rather than a modular component within One?

**Brian Downey:** Yeah, absolutely. Barracuda has been really focusing on the same thing for a long time, which is: How do we provide effective security services for the under-resourced organization?

As we started looking at that two years ago, we saw the platformization happening around security, and we saw the value that people would get through increased information from a single vendor, being able to correlate more information, and also reduce the complexity of managing multiple tools.

That’s what spurred Barracuda One. It was meant to take a lot of the different visibility that we had as an organization and pull that together in a single place.

As we started doing that, we really started honing in on what people want to know about their security and what risks they have. They wanted to understand where their biggest risks were.

What we’ve seen over the last 18 months is AI is one of those biggest risks. Being able to figure out what people are doing with AI, how they’re using it, and helping people with that was a really critical area where we saw a big opportunity and something we could help with.

So it built directly on our Barracuda One story and really started providing different levels of solutions around that. They ranged from things that we saw as features to things that we saw as products.

We recently released some that we put as features within Barracuda One for all of our customers. And what you’re hearing now is that we’ve also released a product set associated with this as well.

**Robert Dutt:** You guys are calling this an industry first, and this is getting to be a hot space, obviously — AI-related security, governance, that whole thing. What are you guys first with here? Is it the capability? Is it who it’s packaged for? A little bit of both? What do you see as the place you’re planting the flag in that regard?

**Brian Downey:** Yeah, absolutely. It’s an intersection of two things.

When we looked at AI security, there are a lot of tools out there, and like I said, we have some natively in Barracuda One to give you visibility.

What we really tried to focus on here was being able to support the entire AI security lifecycle. We wanted to be able to understand what was happening. We wanted to be able to put monitoring controls in place, and then actually allow people to use things and monitor that usage so we could respond if something happens.

That entire security lifecycle view of AI is something that’s very rare.

And when we looked at it in the space that we play in, really having a focus on the under-resourced — ensuring that you’re providing tools that are easy to use and buy and deploy, some of the things Barracuda has been known for within our customer base — it was that intersection that was really unique.

That’s where we really saw it as the first of its kind. We haven’t seen anything else that’s really targeted at our types of customers that also looks at that full AI lifecycle.

**Robert Dutt:** You guys released some research around the same time here, around the same kind of topic matter, talking about 38 percent of mid-market IT leaders today saying their teams lack the skills to secure the AI already in use.

I was actually a little bit surprised the number was that low, quite frankly, given where we’re at with AI and Shadow AI and all that. Curious if you had any surprises in the data as you looked through it.

**Brian Downey:** There were some surprises in the data. Like you said, I was actually more surprised that there were more organizations that had a level of maturity than I expected.

I think that was a really good thing. It was a pleasant surprise for me, seeing that this was front of mind and that people were expecting it.

I’d say through our conversations, though, the thing that really struck me was that even as we got into details around what people were doing, even people who thought they had taken steps to protect themselves, it’s just the overwhelming landscape of AI that usually ended up being the Achilles heel.

We talked to customers who said, “Yep, we saw these risks, so we blocked ChatGPT.” But they’re not even thinking about the thousands of other applications that could have the exact same risk as ChatGPT.

You can start to see the overwhelming nature of the problem in this space. That was a really interesting angle I hadn’t expected: that people would either minimize the scope a bit, or even if they saw the full scope, the overwhelming nature they felt around it.

**Robert Dutt:** Here’s one that I thought was interesting. Your MSP-managed customers said they were more likely to report a skills gap than organizations handling it alone. I think it was around 40 percent versus around 30 percent.

Is that a confession or is that a compliment?

Do you think it’s a matter that MSP-managed customers are more aware of the situation or more exposed?

**Brian Downey:** I think MSP-managed customers are more aware of their own situation.

They’ve already taken the step to get support from an external vendor, being the MSP. So I think they’re aware that they have challenges and gaps, which I think is good.

It’s always good to be aware of where your strengths and weaknesses are, where you need help, and what you want to do internally and externally.

I think that’s where we saw a little more comfort in the admission from some of those customers. I’m not sure the numbers are as different among people who are managing themselves, but I think people might be a little more shy to bring that out as openly as people who have already started working with an MSP as a partner for their IT.

**Robert Dutt:** IBM breach data suggests that almost 70 percent of breached organizations had no AI policies or Shadow AI detection in place.

For an MSP, how do I turn that number into a conversation that SMB customers will, A, listen to and, B, hopefully act upon?

**Brian Downey:** Everything we’re seeing with AI reminds me a lot, in the MSP space, of what we saw in security probably six years ago or something like that.

There was this need. Everyone had this amorphous concern around security, but MSPs had to land that plane and say, “Now this is so pronounced that you need to do something with it.”

Dark web monitoring became one of the key tools in that.

I still remember working with MSPs who were doing dark web scans. I talked to one that was doing them on a golf course. At a charity event, they’d buy the second hole and the tenth hole. They would hand people a report showing them all their dark web risks, and it was just amazingly effective.

I see a very strong similarity here.

When I talk to friends of mine who are business owners or executives at businesses, they’re terrified about AI. And the first question is: They don’t know who’s using it. They have no idea in their organization.

So MSPs being able to come in and shed light on that, drive a productive conversation, help them understand the risks and what that means — what’s good and what’s bad here — I think that’s a huge advisory service that MSPs can provide.

It’s very similar to those early days of security where people had the concern and MSPs were able to embrace them. I really think it strengthened their position with their existing customers and brought a lot of MSPs new customers and new revenue streams.

But I see this as that times ten, if you ask me. It’s something everyone wants to know about. I think MSPs are the perfect people to help educate them on it.

**Robert Dutt:** And with the actions of the Frontier Labs more recently, and OpenAI versus Hugging Face and all that recently, it’s getting a lot of attention and air support. If customers weren’t listening three months ago, they might be listening now.

**Brian Downey:** Exactly. It’s so front of mind, and it’s questions they’re getting from their boards and their stakeholders.

It’s a very in-your-face thing. It’s a very hard one for people to even understand. Like I said, they don’t have their arms around it at all.

**Robert Dutt:** So with the AI Security Suite, a lot of AI security tooling to date has been built and priced for enterprise, which is pretty typical.

I’m curious what you had to change, unlearn or do differently from the enterprise playbook to make this work for the 15-person MSP who’s supporting 200 small customers.

**Brian Downey:** I think it’s a couple of things. Some of it isn’t hard for us because it’s the way we operate normally.

But we had to immediately look at the simplicity of it. We needed to make sure it had the tenancy that MSPs needed. They could look at individual clients; they could look at their whole estates.

We needed to make sure there were simple, easy, exportable reports — something they could physically hand to someone or email to someone.

It bubbled it up into really simple constructs. We didn’t want to spit out a lot of complex technical details. We wanted to make it something that could be understood by someone who was more of a layperson when it came to AI or IT.

Then it came down to whenever we package something, we wanted to make sure you didn’t have to sit there and buy every single Barracuda product just to get this one use case.

The way we were able to keep it really cost-effective, or even free in some cases for some of the capabilities, was by separating them and making them independent.

It allowed you to just adopt our AI services rather than having to adopt all of our security services.

I think that was a really different approach that allowed us to have less friction in using them, but also come in at a more cost-effective place.

**Robert Dutt:** So the pitch is minutes, not months, for deployment, and an experience that’s familiar for those who’ve been used to managing firewalls — which is the vast majority of your partner base, I dare suggest — and, crucially, no AI security expertise required to get started.

That’s a pretty bold claim in a category that tends to beg some degree of specialization.

What’s under the hood that makes that true? And just walk me through what deploying it looks like, sort of day one from the MSP’s experience.

**Brian Downey:** Yeah, absolutely.

If you want to start with our Shadow AI Discovery, it’s literally clicking a button in the UI, and then you have three options.

If you’re using a third-party vendor for DNS filtering, we just ask that you connect that and send information to it so we can pull some of the log information from it.

If you’re using our DNS filters, it just works.

Or if you want to spin up one of our DNS filters, you can actually start a trial right there and use it just for this use case.

So it gives you three different ways you can get that report.

What we did to make the information simple and reduce the AI expertise is pre-populate it with the thousands of common AI applications that are out there. Every single one of them, we’ve put a risk level on. So if confidential information was put into that, what would that mean?

We’re not just telling you what people are using. We can give you a comprehensive view and then summarize it based on risk level.

Everything from that is dumped into a simple report that you can bring to a customer and use to broker the conversation.

All of that would take minutes. It’s just clicks of a button. And again, the output is something simple and easy to understand that brokers the conversation.

**Robert Dutt:** I think that answer may suggest some of this, but let’s talk about the money side of this.

How is this packaged and priced for MSPs, and where’s the margin opportunity for partners?

Is it a per-seat, per-month kind of thing attached across the existing base, a new service line, or is the audit-trail thing more interesting in adding a new type of service that maybe some MSPs have historically given away for free?

**Brian Downey:** What we’ve done is break it down to try to have levels of adoption.

What I just mentioned is available for any Barracuda customer. If you have any Barracuda tool in our Barracuda One platform, you get that capability to get visibility and drive reporting on Shadow AI usage in your organization at no cost.

From there, the next step is to give you simple response capabilities.

We have the ability to start monitoring the major LLMs and looking for information that might be put in that’s confidential, as well as looking for prompts that might be malicious or have other kinds of security risks associated with them.

At that point, we’re talking a very low cost per user, and it’s giving you simple monitoring.

From there, if you want to graduate and have full control over this, you can buy the upgraded package, which provides everything I mentioned but also adds the capability to not just turn AI sites on or off.

So instead of just blocking ChatGPT, you can enable ChatGPT, and every single time a user puts something into ChatGPT, we’ll inspect that package and look for anything that might be proprietary. We’ll look at any text and, if it does contain that information, block it.

So it’s really acting as DLP for an AI tool.

That’s the simple packaging for MSPs.

I look at this as starting out by brokering a really solid conversation with people.

When it comes to end users, it depends on what they’re doing today. I’m talking to a lot of organizations that are using a hatchet and blocking huge swaths of AI tools. That’s a huge productivity loss, and I think business owners understand that.

So I think this is a service that, depending on the MSP’s bundling, they can bring out on a per-user, per-month basis, add it to their services, have an upgraded service with this, and include this as a differentiated capability from other managed service providers.

It gives their customers a ton of value in being able to unlock the power of AI, but do it in a secure manner.

**Robert Dutt:** The audit-trail angle is interesting — the ability to do a one-click export for auditors, for insurance, et cetera.

Do you see compliance, especially where you’ve got PIPEDA obligations, cyber insurance underwriting, local provincial laws that address a given province’s handling of these kinds of things, becoming the actual wedge that gets SMBs to start paying?

**Brian Downey:** I think it’s one of them. I think it’s absolutely a compelling one.

If you’re governed by HIPAA, how do you know that no medical records were put into an LLM that were problematic?

In the tool itself, what we’ve done is pre-build templates so we can look for the associated information with some of the key regulations, like HIPAA.

By selecting that, we’ll start tuning the tool to look for the exact types of packages and information that would be looked at for a HIPAA audit.

And if you click PCI, we’ll do the same thing. Now we’re looking for those 16-digit numeric codes that would be representative of credit card numbers.

So we’ll change what we’re allowing and blocking through these tools based on the regulation you have.

I do think it’s one that’s going to continue to rise as people see audits, and I think it’s a really compelling discussion for MSPs that have regulated customers.

**Robert Dutt:** The skills gap is kind of the whole big premise of your launch, and understandably so. But a lot of times your customers’ customers are looking to buy expertise and not just tools.

And your MSP customers are looking to both secure their customers and up their game.

Can you tell me a little bit about what you guys are doing in terms of training, certification, service enablement — all those things that help an MSP get successful with the product, but also build their broader AI message and game to take to customers?

**Brian Downey:** Yeah, absolutely.

We have a full badging and certification program here at Barracuda, and with the release of AI, we now have those included in that program as well.

So now our users have the ability to get those third-party certifications directly from Barracuda.

I think the other key thing with the skills gap, when I look at MSPs, is the whole goal is to continue to provide higher-value services.

What we’re really trying to do is take out a lot of the legwork that’s required so they can take this information and have a higher-order conversation, a business-driven conversation with their clients.

I think that’s a lot more valuable for the MSP and their clients.

**Robert Dutt:** Barracuda is positioning this as the first of several innovations expanding the AI Security Suite.

I’m not asking you to pre-announce anything here today, but I’m curious what the overall arc on the MSP side looks like, say over the next year or two.

I guess the big one I’m curious about is: Do you see AI governance becoming a standard line item for MSP stacks in the same way that, say, backup or MDR is today?

**Brian Downey:** Yeah, I think that’s the way we’re going.

When we look at it, we’re dealing with what I think is probably the most pronounced problem today around AI, which is, as people use these tools, ensuring that confidential information doesn’t go into them and ensuring that people aren’t hijacking them for the wrong things.

That being said, there’s a lot more.

A lot of MSPs’ clients are going to have tools that have LLMs on their websites. How do we know that no one’s asking that LLM for the CEO’s address or private information from employees?

So now you have to protect the AI that you’re using.

The other one is, I always say with AI, everyone’s a developer now.

I hadn’t developed for a long, long time, and now all of a sudden I’m writing web apps. My web apps probably are over-permissioned. They’re probably chock-full of security risks.

So when we look at the arc of this, I think what we’re doing with AI security is just the start.

People need to start looking at AI as: How do we protect the tools that we’re using and make sure people don’t use AI as an attack vector?

People are going to have to start looking at the tools that they’re building and ask: How do we make sure AI is not introducing more vulnerabilities into our environment?

When you start looking at the directions we’re heading, those are the breadth of the capabilities we’re looking at.

I think there’s a lot more to come here, and I think MSPs will be able to build pretty robust lines of business around it.

**Robert Dutt:** I suspect this may inform the answer here, but I’m curious about the skeptical MSP — the one who says, “Maybe my clients aren’t yet asking for AI governance, so it’s not my problem today.”

What are they not seeing in their own customer base right now? Why should they be paying more attention to this than perhaps they are?

**Brian Downey:** It wouldn’t surprise me if a customer isn’t asking for AI governance. I think most people don’t understand the full risks associated with AI.

So I’d encourage those MSPs to use this as education for their customers.

I was dealing with a family member, and I witnessed them put some contracts into ChatGPT, and I explained to them what that meant by doing that.

I think a lot of people and a lot of organizations are going to be in that mix.

A lot of organizations just have no clue. They don’t think their employees are even using AI, and little do they know they’re using it on a daily basis.

So I’d encourage the skeptical MSPs to just go in and have a conversation.

It’s a simple click of a button. Get the report and have a conversation with your customer about: “How are you guys using AI? Do you know how you’re using AI? Do you have any concerns around governance?”

Have that conversation, and I think you might see their eyes open a little wider than you expected.

**Robert Dutt:** And last one for me. Your research says that organizations using MSPs are the ones who are most worried about this gap, which suggests to me that clients are about to start asking their MSPs, “What can you do for me here?”

What’s maybe the one thing you’d want every MSP listening to do this month to get ahead of their customer on that conversation?

**Brian Downey:** I’d ask them about it.

We saw this in security when MSPs started monetizing security.

The MSPs whose customers had to come to them and say, “Hey, what are you guys doing for security? I’m not really sure if I’m fully secured,” started on the back foot.

They were in a defensive stance, and it was a really challenging one for them to then go in and say, “Hey, the good news is I have this new service, and it’s going to cost you X amount more.”

It was the wrong timing to have that conversation.

What we saw was that the MSPs that were proactive about it, that went in and had these conversations, retained their customers and were able to drive new revenue streams from security.

I think AI is going to be the same way.

So I’d encourage every MSP out there, however you have the conversation, to start having a conversation with your clients about: What are you guys using AI for? Do you know what people are using AI tools for? Do you know what’s going into them? Do you understand the security risks associated with it?

By getting them to start that conversation, I think it sets up an MSP really well to help their clients and to start looking at those new revenue streams they can build on AI.

**Robert Dutt:** So it’s almost a matter of, even if you don’t feel like you are the definitive be-all and end-all expert on AI today, getting out there and having those conversations and starting to build the knowledge of what you need to know to secure your customers going forward.

**Brian Downey:** Yeah. You probably know more than them, so I wouldn’t be too worried.

There are very few people who know everything about AI in this world right now.

I’d still go out, proactively have the conversations, be that business partner, understand their concerns, and that tells you the areas that you want to understand a little more about as well.

**Robert Dutt:** Brilliant advice. Brian, once again, thank you for your time.

**Brian Downey:** All right. Thanks a lot. I appreciate it.

**Robert Dutt:** There you have it, Brian Downey from Barracuda. I’d like to thank Brian for his time, and thank you for listening today, of course.

A few things stood out for me from that conversation.

First, I thought Brian’s comparison between Shadow AI today and dark web monitoring in the early days of managed security was really interesting. For MSPs, simply showing a customer something happening inside their own organization that they didn’t know about can be a very powerful way to start a broader security conversation.

Second, there’s an important distinction here between selling another security tool and creating another managed service. Barracuda is trying to make the discovery and technical work relatively simple, leaving the MSP to provide the higher-value piece: helping customers understand the risk, decide what they should allow, and put appropriate governance around it.

And finally, I think Brian’s advice to MSPs who don’t consider themselves AI experts yet is worth remembering: You probably know more about this than your customers do.

Waiting until a customer comes up to you asking what you’re doing about AI puts you on the defensive. Starting the conversation now gives you the opportunity to become their advisor on the issue before somebody else does.

If you’re enjoying the show, please do follow or subscribe in your podcast app of choice — Apple Podcasts, YouTube, Spotify. We’re on most of the major directories.

And if you’re finding value in it, please do consider leaving a rating and review.

Until next time, I’m Robert Dutt for ChannelBuzz.ca, and I’ll see you in the channel.