Uncover - S37 E8: Pick your side. Pick your story. | The Expert Witness

Episode Date: June 29, 2026

Mosher explains that his Cybercheck technology was built to analyze open-source internet data using AI to generate investigative “intelligence,” not courtroom evidence. Mosher admits the company failed early on to clearly communicate those limits, allowing prosecutors to use Cybercheck as evidence, and he acknowledges the risks of relying on AI without independent verification. However, he defends the system’s value, denies wrongdoing, and attributes criticism to misunderstandings and aggressive defense attorneys. Despite mounting doubts—including inconsistencies in his claims, refusal to share source code, and questions about peer review—Mosher insists Cybercheck remains useful to law enforcement, while the interview leaves unresolved concerns about its reliability, transparency, and continued use. Mosher tells Sam he needs to speak to the people behind the curtain who are having success with it.Binge all 9 episodes of this season on our YouTube page, or get them ad-free on CBC True Crime Premium on Apple Podcasts.A listener's guide to Uncover: Where to go from here.

All right, summer is here. You're out and about. You've got a barbecue to go to or you've got some picnics or, I don't know, maybe a concert. I get it. You are busy. Who's got time to keep up with pop culture? The answer is me. I do.

I've got the time. My name is Elamine, and on my show, commotion, we go beyond the surface of what's new and exciting on television or at the theater or on a bookshelf near you. So if you need help, staying up to date, you can find and follow commotion on YouTube or wherever you get your podcasts. This is a CBC podcast. In the Cybercheck Boardroom, as the air conditioning blew violently upon us, I turned my attention to check on Jeff. He was still there, sitting contentedly like a support dog, as Moser took me through the launch of Cybercheck and those first clients in law enforcement.

We brought clients on from some in California, some in Florida, and some in Colorado. Shout out to the folks we know in Colorado. Some good results, some bad experiences, some great client support, some not so great client support. So it was a mix-mac. We are either loved or we are absolutely hated. Until the agency can see that value add, we're of no use to them. What do you mean when you say can see the value ad? What does this value add to mean? When the match comes back and says that that actually is our guy, that's the value at, they see it.

If Cybercheck pointed to the suspect the agency wanted it to, it's a value ad. If it didn't, it was useless garbage. That's why you have law enforcement agencies that love us. Law enforcement agencies are not going to touch us with a 10-foot pole. The way Mosher explains it, it sounds like he almost became a full-time teacher, taking all his clients by the hand and getting into the weeds of the internet, basically. Judges are still struggling with what an IP address is, and all of a sudden we take all this intelligence and say it's AI driven,

that's a hell of an uphill battle. You need to get not only the law enforcement involved, but prosecutors' offices is involved. It's about providing clarity on what this intelligence is. And this brings us to the first very big issue for Mosher. The whole using cyber check as evidence versus using cyber check as, quote, intelligence on part with, say, an anonymous tip.

It seems like prosecutors and, and police were led to believe that Cybercheck could and maybe even should be treated as evidence. But Mosier does not think that. He explains the mix-up like this. Early on, our company, myself included, didn't do a good job at really driving home how this technology is to be used. I asked Adam about how he addressed this miscommunication. Even in the reports, it says intelligence, you need to do X, Y, Z. It was always there. Moser says if law enforcement officials had read the full report, the label intelligence is clearly stated.

And he's right. On a Derriss Black Cybercheck report, the first Cybercheck conviction, it says, quote, the purpose of the Cybercheck Intelligence briefing is to deliver actionable cyber-based intelligence only, end quote. Moser says that this is the thing that tells prosecutors clearly don't use cybercheck as evidence. So how is it possible that the prosecutors did use cyber check as evidence in both the Salamadee and Adiris Black cases? Law enforcement is busy. Depending on the agency, they could have 100 cases on the go. So page 1, 2, and 3 of the report may not be as deeply read as the cyber profiles of Adam Moser's, I think. dislocation. That's what it's appealing to them. So that's the messaging we had to circle back to

for the people that were using it and we're creating it as evidence because it was clear in the report. Were the sagas that played out in Summit County and Boulder all just reading comprehension errors? Because after the two murder convictions in Summit County, whether the tech was used as intended or not, a new untested technology had been legitimized in a respected court. AI identification of suspects at crime scenes where no human witnesses were present had been embraced. And that was an undeniably big moment for the company. Moser traveling to Akron, being photographed side by side with the prosecutors posing as the new king of crime fighting. It was that success, though, that would attract attention that Moser didn't see coming.

That's the part I was not prepared for, just how good defense attorneys are. When you have a really good defense attorney, they will do whatever they can to ensure their client is either the best deal or the best defense that they possibly can get. And we're referring to the ones in Summit County mainly. The defense attorney there was top-notch. He was brilliant. I'm Sam Mullins. And from CBC's Uncover, this is the expert witness. Episode 8. Pick your side. Pick your story. When Moser and his company found themselves in Don Malarsik's crosshairs, all of their early success was thrown into jeopardy.

Their level of preparation was light years ahead of what we were prepared for in the early days of this. At first, Cyberchek's day-to-day operations were all about interacting with the clients, building relationships, earning trust, schmoozing. But after Milarsik showed up, it was all, subpoenas and thorough scrutiny of everything Mosier had ever done and said. And Moser claims that Milarsik was doing a lot more than just shining a light on Cybercheck's operations. He used the media and he used every resource at his availability to hit us. And he was very successful in doing it.

Going into this interview, I wasn't sure how to broach those critical articles that were written about Cybercheck. But with the mention of Don Malarsik, it seemed as good a time as any. What was it like to have those articles come out just like personally? Like, how did that feel? It's tough. If anybody sits back there and says, oh, it wasn't bothersome. It wasn't, they're an idiot. Like, I took a beating.

Cyberchuk took a beating and I took a beating. Did you regret that you didn't get your own quotes into them? Yeah, we regret that we didn't get our own quotes in there. We always are at the council of prosecutors for the most part. And if they say, you know, our preference is we're not telling you what to do, but our preference would be not to engage in the media, then we have no voice in us. So we don't do that anymore. We always have to have a voice.

Right, right. Hence, Moser is sitting in a room with me now. Again, it's the media and, you know, good, bad, the ugly that got to come with it. Yeah, yeah. I wonder which we are. So, the big pieces come out, Cybercheck has dealt a huge blow, but surprisingly, it wasn't the end. Can you think of a way that it positively affected your company? Sure, gave us clients calling in.

And they're saying, hey, we heard about you guys. One in Florida said, man, you're all over the media, so you must be doing something right for law enforcement. Was their exact words? Wow. Is that cliche about getting bad publicity actually true? Jeff leaned in to finally chime in on this one. It's not the media we would choose. But we're generating a lot of interest.

You know, it's not easy for a small company to generate national interest. But boy, have we? So after surviving the Don Malarsik offensive, Cybercheck was given a second lease on life. They'd learned some valuable lessons that would aid them in the future. The most important of which was they needed to. to be crystal clear in all henceforth situations about how this thing is intended to be used and how it's, quote, intelligence is to be thought of because it's AI, right?

That's the danger. You can never take AI at face value, ever. We are not the means to go kick a door in and arrest somebody in late charges. But ultimately, it is a job of law enforcement and prosecutors to make sure they independently corroborate that. And you've had to change the messaging around it? Big time. How so? Around the intelligence. Because it was being interpreted early on when we first launched at its evidence.

It is not evidence. It is intelligence. Big difference. So you have to be, like, clear about that. Extremely clear. It's surprising to hear Mosher describes product in this way, to essentially say that it's to be taken as seriously as a Ouija board. The Ouija board says what it says.

No one has control over the Ouija board, right? Does the Ouija board know if your crush like likes you? Can it guess the winning lottery numbers? Can it give you the name of the person who pulled the trigger at the murder scene? That's for someone else to determine. You're making technology with such high stakes. Maybe you find someone who is definitely guilty and it's a tip and the police follow up on it. And maybe it's a person.

that was just nearby. Like, how does that weigh on you? It weighs heavily. You can't make a technology like this and it not weigh on you. All of that makes a buffet of chaos. If an AI technology says, hey, Adam Mosier definitely was your guy that did the killing on Friday, April 18. And if no independent corroboration, they go make an arrest on that, charge the guy, he's in prison.

Ten years later, DNA surface that it wasn't him. That's the danger. AI is dangerous when it goes unchecked. We had gone over our agreed-upon time. Do you have anything else that you want me to know about Cybercheck or yourself? Once more, Mosier looked down at the scrawlings in his notebook, which he never once referenced and closed it. No, I'm usually not just open, but you asked some good questions.

So, no, I thought this was, you got everything you need there, Sam, and I thought it was very thorough, so. This was good, Sam, I actually enjoyed it. In the weeks following this conversation, our team continued to learn more about Cybercheck and its reach across America. And thinking back to this conversation, more and more inconsistencies in Adam Mosher's version of events stood out. Like, if it wasn't ever supposed to be evidence, and he knew that the whole time, then why not mention that to prosecutors? And why did Adam participate in the preparation for all of those Daubert hearings if the end goal wasn't? wasn't getting Cybercheck in front of juries.

What about all the other statements in court by Mosier? The peer review, being an expert witness in Canada. What about Cybercheck itself? And those duplicate reports that Don and Marie found, suggesting there might be a human hand behind it all. The more I thought about my conversation with Adam, the more it felt unsatisfying. But sometimes you need a first interview

so that you can do a second one. In a world where crime runs rampant Where mystery lurks around every corner There's one podcast that can save us all Drunk Women Solving Crime Hello, I'm Taylor Glenn And I'm Hannah George And we saw real-life crime cases

With a splash of wine and twist of humor Did you hear about the case of the missing cheese wheel Or the great diamond heist of 2023 Nothing is beyond us Download the latest episode of Drunk Women Solving Crime Where the mysteries are serious But we definitely are I invited Mosier to meet me one month later.

This time in Nova Scotia, his home province. We scheduled the meeting at CBC Halifax. I had a bet with my producer that Moser wouldn't show up alone, that he'd bring a friend with him to sit in on our interview again. And when I met him in the lobby, I was right. This time, he brought Rob Lindsay, the employee who first reached out to us, someone who in person looks like he knows where they keep the weights at the gym.

The vibe felt different. Maybe it was the CBC logo everywhere that helped a penny drop for him. But whatever it was, from the time I shook his hand in the lobby, everything seemed tighter this time. All right. Testing one, two. The first things I wanted to ask you is, yeah, I was looking at the calendar, and it's been like almost exactly a month since we last sat down. So I thought it might be good to just sort of hear what your month has been about.

Generally, the focus remains to be the same is supporting clients, bring clients on, and helping them understand the benefits of the intelligence returned into Cybercheck reports. That's pretty much a day and day out task. Because we are playing somewhat heavier in state. state agencies now. And with that comes a lot of due diligence in their end and things like that. Do these state agencies ask to see the code and do you have to show them the code? No, no. We've never had to show our code. It's always about the results. So how many cases does he currently have running through Cybercheck? We have 132.

132 cases running. 132 cases. The first number that Mosier gave me was a very very specific one. And I guess specific numbers was one of the main things I was here to talk to him about. So I decided this was the moment to get straight into it. In 2022, the Akron Beacon Journal quoted Mosher as saying that Cybercheck had been used by law enforcement 24,000 times in 2,900 cases. And that's obviously a lot of cases for something we now know was so new at the time. 2,900 case runs. Right, right. And what about the 24,000 number? Is that, was that right? That was 2400.

2400 times. Yes. It's been used by law enforcement. Right. Did the Akron Beacon Journal mishear Mosher by a factor of 10? Or did he verbatim say this in court? And I'm reading from the official court transcript here. The lawyer asks, how many either archived or real-time searches have you done for law enforcement since 2017? 24,000. The lawyer replies, 24,000? Mojure replies, yes. I've noticed that with numbers, Moser always has a quick, very specific answer. How many cases are you running right now? 132. How many law enforcement agencies are you working with? 345. All these numbers are known only by Mosier. He doesn't need to show proof of them, and they're always waiting at the tip of his tongue. I've noticed that every once in a while he'll throw out an unbelievable number, like this 24,000 searches one. And if too many people are like, wow, that seems like a lot, he'll walk it back. You misheard me. It's actually 2400. I knew this would likely be my last

interview with Mosier. So rather than get bogged down each time I felt like he wasn't playing it straight, I wanted to make sure that I got to put to him the burningest questions I had. So I moved on to one of the cases that I think about the most. The Megan McDonald case in New York. Daughter of a police officer found dead. No charges ever laid. No breakthroughs for two decades until along comes Cybercheck. And a man named Edward Hawley is arrested and charged. So my burning question was, if this murder happened in 2003 and Cybercheck draws most of its intelligence from things that didn't exist then, like smartphones, modern social media, mainstream Wi-Fi use, Bluetooth devices, etc. Where specifically on the internet, does there

exist any data from our flip phones in 2003? So the internet still existed in 2003. There's still digital traces. And the one you're referring to in New York, we didn't have a large hand in that outside of returning intelligence that they were able to independently corroborate one of the pieces of the phone number. So it's less about a device pinging off of local Wi-Fi or Bluetooth in that case? Right. So what it's all about is we've tracked people's profiles on old phone numbers that they don't even have anymore. It's not always about the device they have in their hand. It's about what? leaves that device, what is communicating with that device, and where that data goes online. Moser tells me it's not always about the devices, but the felony complaint against Edward

Hawley, authored by New York State Police Lieutenant Brad Natalizio, an investigator Michael Corletta says, after conducting an analysis of case information, Cybercheck was able to place the cellular devices of Megan MacDonald, Edward Hawley, and suspect number two together at key locations on the night of the homicide. Hmm. Interesting that. I pressed on. Next, biases.

News coverage of domestic crime in the United States disproportionately features non-white suspects compared to their actual involvement in crime. Given that Cybercheck is trained on these news articles, how do you feel confident that it isn't perpetuating? We are not returning context. We're returning, you know, the IP addresses, email accounts, and things like that. We don't have facial recognition where bias could be implied or actual. Okay, but CyberTech is trained on thousands of articles about crimes. So not using facial recognition and mainly relying on email addresses and IP addresses doesn't mean that it's not biased.

What about confirmation bias? The cops using cyber check might give it names to support their theories, ignoring or dismissing other vital evidence. Because I understand that sometimes police will send you a name of a suspect or names of multiple suspects. What guidance do you give investigators who use cyber check so that they aren't encountering potential issues of confirmation? bias. In terms of an email account belonging to someone, there's not a lot of bias that we can do with that. It either belongs to them or it doesn't. And this brought me to Summit County. In our first interview, Mosher had been open about the significance of Cybercheck being used as evidence in two murder cases. Cases where he now says it shouldn't have been used as evidence.

But if he was the one selling his product to these agencies, and he was the one onboarding them and telling them how it worked and teaching them how it's meant to be used, how could this possibly have happened? We weren't mature enough on the process side of court yet to know the pitfalls. You know, you can't use a report as evidence. You know, even though it says it's intelligence to all this, when it was introduced as evidence, that's when things went sideways very quickly.

Did you tell the police when you onboarded them about, how this is intended to be used? We do, and it's in our reports, that it is intelligence. But again, you can't blame one over the other. I mean, law enforcement, when they get information, they get excited, they want to run with it. Can you tell me how you felt in your initial reaction when you found out that they intended to use it as evidence? Well, when I was there to testify, I testified at two trials there. I wouldn't say we were the only thing they used as evidence, but it wasn't a complete case build where, you know, there was a whole bunch of other evidence introduced.

A lot of our information wasn't properly independently corroborated, things like that. So that was a really big point for us to clean up, to learn from so that it didn't happen again. When you found out that they were using it as evidence, do you go to, like, the prosecutor and be like, hey, We certainly had discussions with them. Can you tell me what those discussions, how they sort of went? Sure, it was, you know, we weren't clear enough when, you know, the report for me and delivered between the police department when they landed over into the prosecutor's office. It was just mixed messages. Do you think that you had the opportunity to really wave your arms and stop it?

No, no. I certainly could have done better at positioning it differently. We put this to prosecutors in Summit County, who, unlike Adam Mosher, did not want to be interviewed for this series. In their statement to us, they said, quote, Mr. Mosier did not have any concerns about testifying at trial about Cybercheck, end quote. They also told us that Moser did not inform them before testifying that the findings of his new AI tool were not intended to be put in front of a jury at a criminal trial. prosecutors did not comment on their interpretation of cybercheck as evidence versus intelligence, or whether for them that distinction even exists.

I checked the clock and realized I was already 90 minutes into my final interview, and I still had so many more questions about what happened in Summit County. When the shit hit the fan, why not just do the one thing that would have ended all the controversy in its tracks, and just showed Don Malarsik all of the source code. Polarsik's seen the tech. It's easy for him to go on the record after the fact, say, he never showed us the tech. Yeah, I did.

The black box isn't a black box. It's the friggin' internet of data. You know, it can be replicated. Why would they still feel like you were withholding? I do not know. You know, when we bring a system down for experts to look through, and then they say, oh, you didn't bring it. down, pick your side and pick your story.

Right. Mosier visited Akron with CyberTech on his laptop in March 24, and months later, Don was still sending Adam subpoenas to try to see the proprietary algorithms, machine learning, and artificial intelligence. Then, as Summit County prosecutors were dropping Cybercheck in the fall of 2024, they co-signed a subpoena with defense attorneys to get Akshack.

access to the source code and data for themselves. Global intelligence refused to provide them with this, too. And some county prosecutors told us in their statement to us that in November 24, Global Intelligence's attorney indicated that it no longer had the source code or data that was used to produce the Cybercheck report. Next, why did you say you were an expert witness all of those times? Why do you think that you didn't tell the truth in that instance? So again, when we talked earlier, Sam was about a miscommunication. Right. That's where it goes to for me, miscommunication.

Right. So every time we were to go into Summit County, I was brought in as an expert witness. So each one of those times I came in and attended their court, I classified that as an expert. However, when I was just sitting there and answering questions, that does not mean you're an expert witness testifying. You're saying that you didn't know what being an expert witness in that context meant? In that context, correct. I see. Yeah.

Okay, right. Next. The peer review. When Moser said Cybercheck had been peer reviewed by the University of Saskatchewan when it clearly had not. We had part of our code looked at by a student out there. Right. And that was part of the development of it.

It wasn't even in the existing code now. So then we said, would this fit peer review? So a student at a university had looked at part of the... the pioneering code, just a part of it, when the thing was originally being cooked up, and Moser wondered, is this what they call a peer review? And then it was up to the prosecutor's office to let it in or not. So it was the prosecutors in Summit County that decided, based on one student seeing part of the old code, that it was good enough to go in front of a judge as a peer-reviewed technology?

In their statement to us, Summit County prosecutors said, quote, Mr. Moser volunteered that Cybercheck was subject to peer review without prompting by the assistant prosecutor who handled his direct examination, end quote.

Okay. What about the incident that suggested perhaps the clearest alternative theory for how Cybercheck actually works? The duplicate Mendoza reports discovered by Don and Marie, suggesting, that there might be a human hand

behind the magic AI machine writing in what the cops wanted to hear. There's two identical cyber check reports that were identical, but there were two different dates. Can you explain how that might happen or what happened there?

At times on what we call triage reports, there'll be patterns of life. That shows a digital footprints of somebody being at or near a location eight or ten days before the homicide and at the same time at the homicide. So you're saying that like someone who has a very rigid routine, that it's possible that.

Stocking behavior. I see. You know, scoping out a bank if you're going to rob it. Right. Going to see digital patterns. The thing that is like hard to believe about that one in particular is just because the murder happened on the second. Yeah.

And this other identical report was for like the 20th. I think you might have a backwards help. I didn't. Oh, you think that the murder happened on the 20th? I think. I see. The murder was on the second, and the report claims that Mendoza returned to the crime scene

and stood at the same place at the same time 18 days after, allegedly killing a toddler. But definitely patterns of life exist for a reason. It was time for us to wrap up. So you've alluded to how there's lots of, Lots of people that have lots of negative things to say about Cybercheck. There's people that will go so far as to call you a con man with a black box technology that he won't show anything to.

What do you have to say to convince them otherwise? I can't convince the defense attorneys that made those comments, right? The defense attorney is there to disprove who I am, who the technology is, who Cybercheck is. That's their job. period. I told you this the first time we talked, Sam. We're in a world where we're either people love us or people hate us. The people that get the data and corroborate the data, they see value in it. The people that struggle with it, defense attorneys and things like that, they don't want anything to do with us. If you had to sort of score it out, like the places

that do see the value in it versus have a struggle to find the value in it. in it? How would that pie chart look like? Percentages-wise, sure. Let's say 50-50. Half the people see the value in it. Yeah. Half the people struggle to see the value in it. Right. It's easy to find people that disbelievers because generally the defense attorneys, for the most part, right?

Right. But it's the people behind the curtains that are using this data, that are seeing every possible value in that data to move a case forward that are behind the curtain. and they're going to stay behind the curtain until they can see their passionate case through. Can you be more specific about the people behind the curtain? Detectives, sergeants, lieutenants,

people whose career are defined by their integrity and value to what they bring to a case. Right, right. Is there anything else that comes to mind that you want to tell me? to like conclude this. No, not really, Sam. Fair enough, I guess. I walked Moser and Rob

back to the lobby. We talked about the Blue Jays. I shook their hands and they were gone. The people behind the curtain. That is an excellent title for a podcast. Going to try to remember that for the next brainstorm.

But seriously, is it possible that Cybercheck is just one of those products where lots of people are enjoying it, but all you hear about are the one-star reviews? Maybe I just haven't looked behind the curtain yet. That's coming up on the final episode of The Expert Witness. So one question. Did you view the Cybercheck report as just a tip that the police were holding back from you? Or were they holding back evidence when it relates to the Cybercheck report specifically?

No, they were holding back evidence, vital evidence. Vital evidence. And they were holding back vital evidence that they didn't even investigate. You've been listening to the expert witness from CBC's Uncover. The series is produced by Raw for CBC. The show was written and hosted by me, Sam Mullins. Our producer is David Waters. The series was developed and reported by David Waters and Jessica Hatcher.

Our editor is Veronica Simmons. Coordinating producer is Emily Canal. Mix by Garrett Tiedeman. At RAW, Deborah Duggan is the head of podcasts. The production executive is Letitia Kiddza-Souza. Special thanks to Emma Wood and Olivia Bhutan. Additional audio from 19 News, 3 News, News 5 Cleveland, CBC News, WKYC, WSOCTV, and WBRZ. At CBC, the executive producers are Cecil Fernandez and Chris Oak.

Tanya Springer is the senior manager and RF Nourani is the director of CBC Podcasts. Tune in next week for an all-new episode of The Expert Witness. Or you can listen ahead to the full series now by subscribing to CBC True Crime Premium on Apple Podcasts, or by subscribing to the CBC True Crime channel on YouTube. Links in the description. Hey guys, Sam here. We'll be back next week with a brand new episode. But between now and then, consider listening to one of the many excellent uncover seasons

that came before the expert witness. My personal favorite is The Village, which is season three. In it, host Justin Ling explores the numerous cases of missing and murdered men in Toronto's gay community dating back to the 1970s. You can find the village wherever you're listening to me now by scrolling back in your uncover feed or by finding the drop-down menu with all the seasons. And make sure to follow us while you're at it. For more CBC podcasts, go to cbc.ca.ca slash podcasts.