AI on Guard: Stopping Gun Violence Before It Starts

Pete Miller [00:36]: Welcome to Predict & Prevent! Today, I’m joined by joined by Tim Fallon, vice president of Strategic Initiatives for ZeroEyes, a company using AI-powered visual gun detection to prevent mass shootings and gun violence. Tim, who is a Marine veteran and intelligence analyst, brings a wealth of experience to the company’s mission-driven role to save lives.

In this episode, we’ll dive into how their technology works — from threat detection to verification and alerting — and explore its applications beyond schools to various sectors of the economy. Tim also discusses ZeroEyes’ unique approach to creating unbiased AI training data and reveals an intriguing development in parametric insurance for firearm incidents. Join us for a compelling conversation about technology’s role in creating safer communities and the future of physical security.

Pete Miller [01:35]: Tim, can you give us a high-level overview of ZeroEyes and sort of what inspired its founding and how it approaches detecting threats and helping to prevent gun violence?

Tim Fallon [01:47]: Well, here’s a stat as a start us off. Last year, 55,000 people in the United States were killed or wounded with firearms, right? So, this is a severe problem. This is a severe peril, right? And it’s been a longstanding one. It’s not just recent. So we have developed a visual AI gun detection capability with a baked in human-in the loop methodology. And what it does is it saves time and saves lives. That’s our mission. And we are company with a mission. So our founders were all military veterans, except for our CTO. And they mostly came out of the special operations world, Navy SEALs. And they went on with life and got their high-powered MBAs, Wharton and such. But they still were looking for some way to serve, right? That was something that they wanted to continue to do.

And there had been a series at that time, school shootings was capped by the Parkland shooting in Florida in 2018. This is where 17 people were killed and 17 were wounded at a high school by a former student who came back with a grudge and a firearm, right? And our founder, our chief founder, Mike, he had kids and they…

At that time, there was more increased emphasis on things such as the school shooting drills and such like that at schools. And his kids were kind of a little disturbed by that, right? And as a lot of kids were, know, and this thought in the back of their head was, are we next, right? And he happened to be at a school then in their gymnasium and he saw that there were cameras in the gymnasium. And he asked the school resource officer who was on duty, like, is anyone watching these?

And the answer was no, right? These are basically a forensics tool. And that’s really where the company was born because he said, you know, look at all these cameras out here. Look at this capability that we could have, but we’re missing, right? So, and that’s how we developed the video analytics that is going to be looking all the time for a weapon. And that’s what’s going to be getting the information to people to better save time and save lives.

Pete Miller [04:00]: And how did you get involved, Tim?

Tim Fallon [04:04]: I came up through the military, right? I was in the US Marines. I was in Afghanistan. I actually was wounded there pretty severely and I had to give up the military life. But I went over to the intelligence side of the house working as a civilian intelligence analyst for Department of Defense. And I spent almost 10 years there and I kind of wanted to get out and see the rest of the world, right? See the real economy.

And a friend of a friend got me and contacted. And when I saw what ZeroEyes was doing, I immediately fell in love. And I thought, man, look at this mission, at what these guys are doing every day to help people. And it really felt like I was returning to the military. That daily mission, that daily movement of let’s get things done because we have a bigger mission, a bigger issue to fulfill. And I’ve been there for two years now and been loving it.

Pete Miller [04:54]: You know, Tim, I think a lot of people in our industry can relate to that. Maybe not at the same level, but I think the insurance industry makes people’s lives better and safer, right? So, I think there’s a real sort of overlap between what ZeroEyes is trying to do and what the folks that we talk to in the industry. So, like walk us through how ZeroEyes, like there’s the camera and you’re in a school. How does it work? For example, like what happens with school and law enforcement or just kind of the process when an alert is triggered. Like how does that whole process work?

Tim Fallon [05:28]: It’s a three-step process, right? So it’s detection, verification, and alert. So our capability, our algorithm sits on a customer’s existing camera infrastructure, so long as you have IP capable cameras. And if they’ve been installed in last 12 years or so, it’s probably something we can work with. We don’t work with analog and such. Right? But as long as it’s that kind of cable camera, we can work with it. We’re camera agnostic. And all it does is it sits there.

and asks itself frame by frame, looking at the camera feed, is there a gun in this image? Is there a gun in this image? It’s over and over again, right? Is there a gun in this image? And when it believes that it sees a gun, that’s the detection, right? And then we move over to verification phase. It’s flagged it and it put it in front of our folks.

Now we have two 24 seven operations centers that we own and manage them, we do all the hiring for them and the training. And one is in Philadelphia in Conshohocken and the other is in Honolulu. These are open 24 seven and they’re staffed by foreign military, foreign police, right? So these guys know the stress of these kinds of jobs. They know how to conduct themselves on a long watch like this. And they know how to identify these issues, these kinds of things in a very timely manner. So in as little as three to five seconds, they can identify whether or not it’s a false positive, someone with an umbrella that our AI picked up, or if it’s a real deal, someone with a weapon and with malicious intent, obvious malicious intent, intent to cause harm.

And then we can move to the alert phase. So simultaneously, our analysts will contact the customer’s pre -designated contacts, security contacts for such a scenario, as well as first responders, so police, medical, fire, rescue. And we’ll be providing all these folks with, again, simultaneously with an image, a description of the perpetrator and the event, as well as a point on a map. So this gets the right people to the right place at the right time.

And we can deescalate and escalate as necessary based on what our analysts are seeing.

Pete Miller [07:29]:

So that’s very interesting. you know what I was wondering Tim, it’s like, there’s a lot of states that have like concealed carry laws, So it’s the determination of the threat. Can you just dig into that a little bit more? Like once it’s been detected, but like you’re not sure, is that just their training in the military or like, how do you decide that that’s a threat and you should forward the information?

Tim Fallon [07:58]:

Well, that falls on our folks at the 24/7 ops center, but these guys are highly trained and they’re also co-located, right? None of this is done remotely. So, they’re in a room with their supervisors who are more experienced with them. And if they want a second opinion, all they have to do is ask for it. Instantly, there’s someone else looking at it, giving a second opinion and moving on that escalation. And to your point about, states are different with their laws regarding firearms, right?

So, we know that and we can see when a tagged image comes up that the algorithm has done the detection work on, we can see where this is geographically. So, we know that seeing a weapon in a parking lot in Texas is potentially very different in how we react and what we’re doing than if we saw it in New Jersey, say. With very different gun laws. But the thing is, and in both cases, we can still contact these customers directly, whether we contact first responders as well, or we’re just contacting the customer to give them a heads up on what’s going on.

Pete Miller [09:04]:

Can give us an idea, you know, ZeroEyes, how many instances of, you mentioned 55 ,000, so gun -involved violence are occurring, so I understand that, but what about at schools and how many events have you guys flagged and created alerts for?

Tim Fallon [09:23]: Right, so last year, last academic year, 2023-24, 35 people were killed on K -12 campuses in the United States, and 123 were wounded, right? And that’s the data that ZeroEyes has. So this is a grotesque problem, right? And what we are able to do, as I said, is to provide that information to help be a proactive capability, right? See the weapon before a shot is fired and be able to provide information, live intelligence to first responders and security folks during an incident, right? And since we’ve gone live, we’ve had over 1,000 instances where we’ve been contacted by customers and first responders. And remember that de-escalatory capability, right? So sometimes we might see, say, an airsoft gun on the playground. And we can identify that as such.

So rather than having to call the police and trigger a lockdown and other traumatic event like that, being called to school directly and inform them on what’s happening. And the SRO or the principal can go down and take care of that problem as they see fit, right? So we have a DSO to our capability. We’ve also led directly to 12 arrests and on a weekly basis now we’re having pretty positive impact direct with police.

Pete Miller [10:39]: You know, even saving one life, right, of a child, that’s amazing.

Tim Fallon [10:43]: And that’s the mission, right? As we discussed, this is a mission -oriented group, and that’s the mission we wake up for every day as a safe people.

Pete Miller [10:50]: Yeah. Thank you. mean, that’s awesome. So beyond the company’s focus on schools, what other kind of customers could be? I mean, people in the insurance industry underwrites basically society. So there’s a lot of people in our industry. What other kind of customers does ZeroEyes think that they can, where this is applicable?

Tim Fallon [11:09]: Yeah, our traction is excellent, right? And it really is across the economy that we’re getting customers because there’s interest in it. And it really comes down to the fact that if you have a public facing entity or even a workplace, this peril is here, right? This peril is out there. So, we have about 50% of our clientele are K through 12 schools in the United States.

The other 50 % is everyone from the Department of Defense to major US transit systems to manufacturing to retail. You know, those first two, DOD, as well as transit, we can talk a little bit more about those. So transit, there’s a lot of throughput on that, right? And lot of public access, obviously, but it’s a critical infrastructure. The buses need to run, the trains need to run. Otherwise, the economy in a major city kind of grinds to a halt. And we can help support that by making it a less dangerous place and being able to ensure that the police services that furnish and protect those transit services have good information in such an event, right? So we can cut through the fog of war, cut through the chaos, be a proactive measure and feed them the right intelligence they need.

Also Department of Defense, we work with them on specific issues, right? So we actually have a wholly owned subsidiary, ZeroEyes Government Solutions, ZEGS we call it. And they’ll work with the Department of Defense to do other detections other than just firearms, right? Really specific requirements that they have. And that’s been really successful, right? But all this is to say that really is across the economy because this is a peril that is also across the economy.

Pete Miller [12:53]: So if the typical customer, you mentioned the founder was sitting in the school and saw the camera, what would the typical customer have in place for you to be able to work with them?

Tim Fallon [13:05]: Again, all they need are IP capable cameras. So internet capable cameras, right? And they can’t be, they have to be a digital camera, can’t be an analog camera. And we do have floors on the capability of the camera. So 720 or 1080, depending on the situation, right? But again, a good rule of thumb is if it was bought in the past 10 or 12 years, we can work on it. And that’s all they need. That’s all a customer needs. We can do now, actually very recently, we can do as few cameras as one, right? We have a new capability, ZeroEyes one, where customers that just want one to 16 cameras can buy an appliance, right? A lot of the work is actually done online, although we do the final testing on the install side of the house. But then once they’re on there, it goes to the same 24 seven operations center that our huge enterprise clients go to, right? So we can do everyone from one camera to thousands of.

Pete Miller [14:03]: That’s really interesting. Can you just dig a little more into the technology? So you talked about it a little bit, can you just, for the folks who listen to this, can you say, you know, what, you know, a little more on the technology that is deployed and what, you know, what makes your solution different from other firearms detection solutions?

Tim Fallon [14:021]: really, simplistically, it’s just computer vision. It’s just computer vision object detection. And that’s all it is. It’s obviously extraordinarily advanced. And we have our dev and AI team. I understand every fourth word they say, quite honestly. But it’s computer vision. That’s what the AI is, right? And what sets us apart from our competitors, though, I think, are three things, right? So it’s our leadership, our scalability, and our adherence and recognition of standards.

So, leadership in the field, we are one of the first movers and our methodology has been, has really proven us to be better than everyone else. So, when you’re talking about human in the loop, our organic data capabilities that we do, the training that we do, how we keep that in house, that’s really starting to become a standard and we’re the ones that are leading that.

And then there is also the scalability, right? I already talked about it, but we’re on tens of thousands of cameras in over 43 states, hundreds of customers, and it’s growing every day. And then we’ve shown that at scale, we can operate this. Everything from our training of the data to the human in the loop to the alert functions, we can do this at scale.

And then thirdly, what says so far is adherence standards, right? We look for, seek out and adhere to our various regulatory and the standard situation formats out there. And this is everything from HIPAA to ISO 27001, SOC 2, also DHS, Federal DHS Safety Act designation and vetting. And really what this allows us to do is be easily and reliably be a partner for enterprise customers.

Pete Miller [16:09]: You know, one of the things that sometimes computer vision, particularly AI computer vision, there’s bias. So how do you make sure that the model is not biased?

Tim Fallon [16:19]:

Yeah, great question. And ours is not biased because we do no behavioral analytics, right? So we’re not looking for anything to do with the human body and how, and also we don’t do anything facial and nothing like that, right? We look exclusively for the weapon. And when we actually train the data, we actually will, our analysts who help create the actual training data, well, some will be wearing gloves at times, right? They’re different colors. So, it doesn’t matter what color your hand is that’s holding the weapon, whether it’s white or black or green or purple or whatever, right? It’s all our capability is doing is looking for the weapon, right? So, there’s no bias in it because we’re just looking for a brandished firearm.

Pete Miller [17:03]: Interesting, because I know, I mean, that’s a big issue and justifier, I’m certainly, as you say, it’s not behavior, it’s not color, it’s not anything else. So the companies, I understand the company’s created its own training data sets. Can you tell us a little bit more about what that specifically means? And again, how is that an advantage over training AI on third party data?

Tim Fallon [17:29]: Right, so there’s several things there, right? But first, how we do this. We built from the ground up all of our training data ourselves, right? Because you can pull, you can do internet scraping for all sorts of gun imagery, but a lot of that is marketing material, firearms marketing material, right? And what you see in marketing material is not what a security camera will see at night, right? So, we need to build the data in a realistic way.

And how we did that was just doing our own imagery. So, hundreds of thousands of images of our own analysts holding weapons in different positions, different kinds of firearms. And then later, this is what we use now, we have our own AI lab, right? And that’s also in Conshohocken. Big green screen room and we have hundreds of different kinds of commercially available cameras plugged into the ceiling, and we can take hundreds and hundreds of images of people in the green screen lab. what was important there is that we can change the background, right? So, we can change the background to a casino floor or to a beach or to a busy street, right? And we’re making sure that we can find that firearm no matter what the position is and no matter how busy that street is, no matter what camera is on, no matter what the angle is, it’s all really critical stuff. And so, what is important about this is that because we built all the training data, we know exactly what went into it. And we have a supreme confidence in what we’re pulling out of it. And that is really what sets us apart on the training data side of the house.

Pete Miller [19:08]: I understand you were recently participated in Lloyd’s of London accelerator program for tech startups. So how does that program help ZeroEyes? Particularly look for opportunities in insurance?

Tim Fallon [19:25]: Yeah, so, and you mentioned earlier about how with insurers really helping people, right? That’s what they do. But we recognized pretty early on that, you know, there is a relationship there that we are both insurers and ZeroEyes are with their customers on their most catastrophic days, right? And we want to be able to bridge that gap. And that’s what Lloyd’s was able to help us with. Now, the Lloyd’s Lab program, they run it out of London, obviously. It’s really fantastic. I can’t speak highly enough about it. ZeroEyes, we came in not knowing that much about the ins and out of insurance and the mentorship program they were able to help us with and just the baseline information they were able to help us with and building us up toward product completion was unbelievable. And it’s an ongoing relationship, right? So we’re done with the accelerator, but it’s a lifelong now relationship that we have, ZeroEyes has with Lloyds. Pretty amazing.

But some of the things that we were able to build out was, how do we partner with insurers, with risk managers, with carriers and brokers? What can we offer them? Instead of us trying to sell ourselves through insurers, what can we offer them? What kind of insurance actually can we embed into our capabilities, our products? And one of the things that we’re thinking, that we’re working on, or actually working on with a major London syndicate, is a firearm parametric.

And to give you an example of what we’re trying to cover here, is that in the average high school in the United States, even if a weapon shows up on campus, there’s going to be costs associated with that event, regardless of the outcome, right? So, let’s think of it this way. Even on ZeroEyes most successful day on a school campus, right? We identify our firearm before shots are fired, we’re able to vector police and security to that firearm, and they are able to interdict it and that perpetrator before a single shot is fired and before anyone is hurt.

Still in that optimal case, an average US high school is going to pay in the five figures of unanticipated costs associated with that event. And these are things from business interruption kind of costs to employee issue kind of costs to medical costs to property damage. runs the gamut. It’s going to probabilistically in the mid -five figures. And we thought you know, this is something that we can embed into our product and make ourselves a more holistic product, right? So not only are we saving lives, right? And that is, and will always remain the most important of our capabilities, right? But we can actually start to be able to help people financially that are trying to deal with this kind of situation. Right? So it’s opened up this whole new world of what we can be doing to help our clients and to help the insurance industry protect people from this kind of violence.

Pete Miller [22:13]: Yeah, that’s a really interesting application of parametric. I would not have thought it. I didn’t know either. It was, you know, five figures. That’s super interesting. That’s a very, very good idea. Yeah. I like that. So, fascinating product. So Tim, as you look forward kind of into the future, can you shed some light on plans that ZeroEyes has and where you see other opportunities?

Tim Fallon [22:40]: Yeah, so this problem, this peril isn’t going anywhere, right? Look against other crises, right? So Y2K came and went and COVID came and went, right? But gun violence is here to stay for a host of reasons, right? And we really just think that our future is going to look something along the lines of what happened to airbags in the 80s and 90s and fire prevention for buildings in the 60s and 70s.

we want to be the next smoke alarm, because that’s a detection capability. And before that, building’s going to be much more easily burned down, just because people just didn’t have the situational awareness of what was happening, where smoke and fire was, and what the problem was. And weapons detection and visual AI weapons detection, I think, most specifically, has that ability to be a critical layer.

Not the only one, mind you, but a critical layer, just like a smoke detector is a critical layer in fire prevention. So for instance, there’s about 100 million cameras, according to some estimates in this country right now, security cameras, there’s probably going to be 200 million by the end of the decade. So, the more cameras that we’re on, the more people we’re going to save. So we think that the future is going to be in that direction.

Pete Miller [23:59]: So, I like to ask folks, Tim, I want to do a thought experiment. It’s 10 years from now and ZeroEyes  achieved its goals. What has gone well? What have you guys done right to kind of, first of all, what does that look like to you? And then what have you done right to meet that future?

Tim Fallon [24:17]: Well, a lot of it, I keep hitting on this one, but the more cameras are on, the more people that we’re going to be helping. That’s the mission. And we see ourselves on millions of cameras. That’s success. And that we’re actually able to open up into more products in this realm. So, like our potential parametric product and maybe perhaps even, far down the road, of course, but other analytics that we could bring to the table.

But what brings us to this point is our ability and willingness to adapt and to find new ways. So, for instance, this whole area of gun violence has been plagued by decades and decades of debate and debate and debate. But we’re just trying to come to the table with a solution to help people now. And that’s been showing our flexibility, our adaptability in building out a blue water market.

And very successfully, right? And I think that’s what really gets us there is our willingness to adapt, keeping our eye on the mission and our willingness to take effectiveness over efficiency when we’re putting in the resources toward solving problems.

Pete Miller [25:29]: Tim, you know, I’ve heard before from ZeroEyes that there’s sort of a multi-layered approach to this issue. So where does ZeroEyes fit in that?

Tim Fallon [25:37]: We stand at a point where physical security really is changing for a host of reasons right and ZeroEyes I see us as a critical layer but only one layer in it. So when you are an organization whether you’re a school or a retailer or your manufacturer or whatever This is a good time, these days are a good time to take a good look at your security, how you’ve set it up, what layers you have in there, and what can be changed, right? Really, where are your threats, right? And again, I see us as a critical layer, but not the only layer. And there’s a lot of good work being done out there, not just ZeroEyes

Pete Miller [26:14]: So if somebody wanted to find out more about ZeroEyes, how could they do that?

Tim Fallon [26:18]: Well, there’s just obviously the website. And then we have a lot of LinkedIn capabilities. So you reach out to any of the sales folks or folks there. on the insurance side of the house, I’ve been kind of on the strategy team. I’ve been leading the horse there. So I’d be willing to talk to anyone there. I’m easy to find me on LinkedIn.

Pete Miller [26:40]: Well, Tim, I think this is fascinating and, you know, thank you for your service when you were in the Marines and thank you for your ongoing service. From my perspective, I think, you know, we at the Institute are a mission-driven organization. I believe insurance is a mission -driven industry and I applaud you and ZeroEyes for your focus on the mission. And I can’t think of a better mission than saving young people’s lives. You know, I mean, that’s incredible.

Thank you very much. Appreciate your time. And I know there’ll be a lot of people in our industry interested in this.

Tim Fallon [27:15]: Yeah, no, thanks for your time. Yeah, always willing to help. I love this industry. I’ve been really enjoying getting to know people, but it’s just as you say, yeah, it’s really mission focused and what can we do to help people? What can we do to reduce risks and help people?