The Nancy Guthrie Case Is…Nearly Solved?

The Nancy Guthrie Case Is…Nearly Solved?

The criminal or criminals who abducted Nancy Guthrie left behind a trail of evidence which has been collected by authorities—authorities who are now using artificial intelligence to utilize those data sets to further the investigation. Today we are talking to Morgan Wright, CEO and founder of the National Center for Open and Unsolved Cases, who walks us through exactly how that is going down.

Welcome to In the Know. I am your host, Jackson Buhigh. Joining me today is Morgan Wright, CEO and founder of the National Center for Open and Unsolved Cases, and also the founder of the Crime Reconstructed podcast on YouTube and Substack. Morgan, thank you so much for joining me today.

“Jackson, hey, it’s great to be back with you again.”

Of course. Now, we have officially crossed the four-month mark since Nancy Guthrie was initially abducted. And you have been going around to a variety of media outlets explaining what the next step is in this investigation. Can you tell our audience a little bit about that?

“Yeah, you know, this actually came out of CrimeCon. I got approached by Fox Digital saying, ‘Hey, look, there was some information that the FBI may have new tech.’ They were not exactly specific about it. So I said, ‘Well, look, let’s think about this for a little bit. What could some of this new tech be?’ So that is where it came from.”

He paused, gathering his thoughts. “I thought about this for a while. I did quite a bit of analysis. And this goes back to some lessons I actually learned from a friend of mine who was the head of the director of science and technology for the CIA. They did real big stuff. They hunted high-value targets. We looked at patterns of life, signals analysis—where somebody would go, what would they do—and then track that electronically.”

“So I think probably one of the highest probabilities that they have had a breakthrough is the application of AI to be able to look at these massive data sets. We are looking at cell site information. We are looking at adtech data. Adtech is all the data that comes from the developers collecting from your mobile apps. That is independent of the carrier.”

Jackson leaned forward. “And you applied this to the Guthrie case?”

“I actually applied that to the Guthrie case while I was out there. We were doing signals analysis. But it is about taking all of that together—with vehicle traffic, with videos, with everything—and starting to put it together and asking questions of it. ‘Here is the pattern of activity I am looking for. Show me the person. Show me the pattern of activity that indicates somebody did pre-reconnaissance surveillance, say back in January.’ And then you see something come in.”

Morgan’s voice took on a sharper edge. “I am going to say this is kind of like the Kohberger case. You are going to have something come in. They know that some of this can be tracked, so they are going to shut off their phone at some point. But then it is going to reactivate. Well, look, Jackson, that is a tell, too. You shut something off, you turn it back on. For me, that is a red flag. Why did you shut it off in the first place?”

“So I think there is going to be a signals analysis breakthrough with AI where they look at all of the stuff. They are pulling in license plate readers. They are pulling in video from all of the different homes that were out there. They are pulling in video from the businesses. And what they are looking at says, ‘Let us bring this together. Let us synthesize this together. Show me the vehicles that were engaged in a pattern of activity that are consistent with pre-reconnaissance activity or consistent with a targeted abduction.'”

“And I am saying it is a targeted abduction. I will tell you why really quickly, because this goes into why I think this is the leading candidate. Everybody for a long time, Jackson, talked about, ‘Well, why did the guy take the camera?’ I mean, he was trying to disguise himself. No. The porch guy was absolutely comfortable being on that porch in his disguise because he did not think he would be identified. He was there several minutes. We know that. He tried to use vegetation to cover it up. But why would you take the camera?”

Morgan leaned forward. “You take the camera because he is not worried about disguising himself. He is worried about disguising the vehicle. And if we know precisely what kind of vehicle is out there, I can now apply all of this other analysis to say, ‘Hey, it is a white Honda.’ And now I can follow that white Honda everywhere. I can trace it back.”

Jackson nodded slowly. “These are lessons we learned from Afghanistan and Iraq?”

“Exactly. I had a friend who worked at DARPA—the Defense Advanced Research Projects Agency—and they built this thing called Gorgon Stare. It collected tons of high-definition video of everything. It was just loitering in the area. So when they would have an IED, what would they do? They would find it, see who was around it, see what happened before it, and trace the vehicle back and the people back. So this is very much about transportation, Jackson. So I think the biggest breakthrough—and I do not know this, this is all just hypothesis—but the biggest breakthrough I think is going to come from some type of forensics on the signals analysis, cell site and adtech data, and being able to combine that with the video to determine: do we have something that is indicative of a pattern of activity consistent with a targeted abduction?”

Jackson asked, “Now, I am glad you bring up data sets, because I am wondering: what if this person knew exactly what they were doing to the point where they could get around things like CCTV cameras and having their cell phone ping a cell tower?”

Morgan shook his head. “You know, a lot of people do not realize, but there are a lot of nature cameras out in that area. Not all of them face the road, but there are some on the road. And there is actually part of that—the Sonoran Basin, I think they call it—it is a preservation area. But you have also got traffic cameras. The Arizona Department of Transportation had traffic cameras out there. You have license plate readers. I do not know exactly how much they got, but to your point, let me tell you—it is about data sets.”

His voice dropped slightly. “I did some classified work that I was able then to apply in an unclassified setting to work on the D.C. Sniper case. I predicted a model that said the D.C. Sniper will have already had contact with law enforcement multiple times. He had been in—the tag on the vehicle had been checked by law enforcement multiple times. So a lot of this is about: we have got the capability now to combine large data sets, have the ability for AI to run models against that, to say what is most likely.”

He spread his hands. “Again, we do not know what the FBI has got. We do not know what Pima County has got in terms of all the data sets. But there is more data out there than we know about. But here is the other thing, too, Jackson. I could be looking at something like whether they knew the area very well and they stayed outside of where cameras were. What does that indicate? That indicates pre-surveillance reconnaissance. It is what we have identified in terrorist planning cycles.”

“They do not know they are doing it, but part of it is broad target selection, initial reconnaissance, specific target selection, rehearsals, action on the objective. So there are things that they go through. People that do this go through specific steps that behaviorally we have seen over and over. So it is applying behavior models into this as well, to say, ‘If we have a behavior like this, what are we looking for?'”

Morgan leaned back. “You and I talked earlier. I have friends that worked from New Scotland Yard on the 7/7 train bombings. When you go back and look at the four attackers, you know what they did a week before? They all got together with their backpacks. They met at the train station and went their separate ways. They did exactly behaviorally what attackers do in things like this. So I think that is what we start looking for. We look for behavioral things. We find the data sets and start applying AI to them.”

“Here is what AI will do. It is not that we could not do it if we did it on our own. It is that it can get to that solution a thousand times faster than we can and create additional opportunities, additional leads. Because right now, one of the most precious resources we have is time. And as time marches on, some evidence will tend to disappear over time. And that includes recording over cloud recordings that you might have or anything like that. So time is the key thing here. AI accelerates our ability to get to critical information, critical evidence—quicker, faster, sooner.”

Jackson asked, “AI in the grand scheme of things is fairly young. I am wondering, in your experience, if you have ever seen a case where AI has been implemented to solve an open case.”

“That is a good question.” Morgan sat up straighter. “My work at the National Center for Open and Unsolved Cases—we are actually working on what we call Version Two of HOLMES. HOLMES 2.0, named after Sherlock Holmes, the world’s first private consulting detective. We are intentionally applying AI to the investigation of criminal cases. I have these things called structured prompts. We have actually worked with the Texas Rangers. I have worked with the U.S. Marshals. But it is to take AI, take all this information and feed it into it, and based on publicly available information, create a structured prompt that allows me to look at—I can input cell site information, I can input adtech data, I can input video.”

“But on one of the cases—I cannot say which one yet—but on one of the cases, the U.S. Marshals’ top fifteen, I have been able to identify what are safe houses. We know because of pattern activity: they are at a place for one week, they move, they go to another house for a week, they move, they go to another house for a week. That is behavior indicative of safe house activity.”

Morgan’s eyes were steady. “I think AI has been used in certain areas. One of the things we are working on too is to provide that training and guidance before the courts do. Jackson, I will tell you—if we wait for the courts to decide how law enforcement can use AI, it is a losing proposition. So we have to get out in front of it. There is one thing about making AI safe and strong and good—that is like creating a safe car. But what we are talking about is the rules of the road. How do we teach the drivers? What do we teach them? That is what we are focused on. How to recognize bias, how to recognize hallucinations, how to recognize structural issues in your prompt.”

He paused. “So I think AI has been used, but I do not think it has been used at scale yet to have the kind of impact we are looking for. Because as you point out, it is relatively new. I mean, two years ago ChatGPT came out. So we are still in the—we do not have any idea where it is going to take us, but there are a lot of solutions being built around AI right now.”

Jackson pressed further. “Now, if you gather evidence using this AI software, I am just wondering: if you gather evidence on this person and then you hand it over to the prosecutor, is that something that is admissible in court, or do you have to dig a little bit deeper to find real, rock-solid evidence?”

“Well, that is a great question, because there are certain things you can do.” Morgan’s tone became more measured. “One of the things we are doing, at least with our prompts, is complete transparency. I use a combination between Claude on one side and Grok on the other. They both do fabulous things. Claude creates everything, feeds it, injects it into Grok. If you techies out there, it is called via MCP connection—Model Context Protocol—but I can actually run it on a Chrome browser. But I capture everything we have written from the cloud side. I capture the exact inputs into Grok on the other side.”

“And so, as long as it does not make the decision for you—because that is the issue with some states—a lot of it is not that the AI is admissible, it is that your evidence is admissible. Did you legally collect it? Right now, what you are talking about is, in a sense, if I did a reconstruction of something like the mask on the intruder, if I were able to reveal that mask and say, ‘Now we know who it is, here is his face’—that is going to have to undergo, I think, a process they call the Frye test.”

He held up a hand. “That is why you can admit DNA. That is why you can admit fingerprints. These are things that have been scientifically validated, are repeatable. There is a general consensus in the scientific community. But that is why polygraphs—I used to teach at NSA. I taught behavior analysis, interview and interrogation. Used polygraphs a lot. But Jackson, polygraphs are not admissible in court because they cannot pass that Frye test.”

“So I think certain aspects of DNA, if we have new tools to deconvolute DNA, that has got to go through scientific testing. So I think as long as you can trace it back to say, ‘Look, I still made the independent decision that we have probable cause to do X,’ I think that is going to be admissible. But I will tell you this from experience: you have got to be very transparent about what you did. Because if you are not, and if there are holes in your case, I guarantee you who will find them will be the defense attorney.”

Jackson nodded thoughtfully. “You have walked us through some of the specific data sets that this AI software would utilize to further the investigation. But if you were spearheading the investigation personally, what would you be doing specifically with this AI software you speak of?”

Morgan’s eyes lit up. “Wow, that is a good question. You know, one of the things I would use it to do—I would spend a lot of time. Einstein said one time that if you have one hour to solve a problem and your life depended on it, spend the first fifty minutes defining the right questions to ask. I think it is something close to that. Do not hold me to that, folks. But it is something very similar.”

“So I would spend a lot of time really defining what are our boundaries, what does our structured prompt look like, have we written the proper what we call skills in there that can do certain things for us. So when I spin up Grok, I have got four agents running in Grok. Each one has a separate task. So on this, I would look at—I will give you an example, Jackson. Perfect case. You guys have covered it. It has come out: the Long Island killer, Rex Heuermann.”

Morgan leaned forward. “The evidence that led to him, people said, ‘Well, DNA solved the case.’ I said, ‘No, DNA did not solve the case.’ The original lead about his Chevrolet Avalanche is what solved the case. And it only solved it when they put a task force together, went back and relooked at the data, then combined it with cell phone data and were able to develop a lead—which then they developed Rex Heuermann as a suspect, got his DNA off of a pizza crust, and then identified him as the Long Island killer.”

“So DNA did not solve the case. Going back and doing the hard work—and this is where AI can accelerate that—can find all the things that you missed, the non-obvious relationships. So that is what I would do. I would make sure that we have everything structured, and then I would start feeding it in to say, ‘What other opportunities are there? What did we miss? What else could we be doing with this?'”

He spread his hands. “I would feed in—I would look at the ability to put in large amounts of data, of video, of cell site information. Anything that they lawfully—legally collect—I would put into this. But I would be very careful about how we structure it. So that is what I would do. And what I would do is go back and not say, ‘Hey, we do not trust what you did.’ But what we have to say is that, ‘Look, you did what you did based upon the initial information you had, your initial mindset.'”

“Part of the problem with working cases, Jackson, is we have these things called heuristics. We want shortcuts in our mind. We want the most logical—what logically explains things. What makes sense to us. Well, it is not always about making sense, because we skip from evidence past the proper interpretation and go to narrative. So AI can also help you eliminate some of your own bias in there and expose those things.”

His voice grew softer. “So for me, I would line up everything. I would identify where are our gaps of knowledge. Do we have anything that can close those gaps? And I would take nothing for granted. I would say, ‘We have already looked at that. I do not care if we have already looked at that. We are going to go back through and recheck everything.’ And at the end, I want an output that says: what is known, what is knowable, what can we never know.”

He ticked off the points on his fingers. “I would have buckets of information that I would break it into, and then I would assign tasks to go fill those buckets.”

Jackson smiled. “You can find Morgan Wright on the Crime Reconstructed YouTube channel and Substack. Mr. Wright, you have been absolutely amazing. Thanks so much for joining me today.”

“Jackson, it was always great, and I appreciate it. It is always good to have these kinds of conversations. So thank you.”

The screen faded to black, but the questions lingered. Nancy Guthrie has been missing for over four months. The person who took her knew enough to wear a disguise, to snatch the camera, to try to hide a vehicle. But no one disappears completely. Not in an age of license plate readers, traffic cameras, nature cams, cell site data, and adtech tracking from every app on every phone.

Not when AI can sort through millions of data points in hours instead of years.

The evidence exists. Somewhere in the vast digital ocean of information that surrounds every modern life, the pattern is waiting. The white Honda that appeared and disappeared. The phone that powered down and powered back up. The face behind the mask, revealed not by a single lucky break but by the slow, relentless mathematics of probability.

Morgan Wright built his career on cases exactly like this one. He helped hunt terrorists. He helped map the behavioral patterns of serial killers. He helped the D.C. Sniper task force understand that the shooters had already been stopped by police multiple times before they were caught.

He is not guessing. He is applying lessons learned in war zones and classified intelligence operations to a quiet neighborhood in Arizona where a woman vanished from her own porch.

The investigation is not cold. It is processing. It is synthesizing. It is waiting for the AI to finish what human investigators started.

And when it does, the question will not be if they find whoever took Nancy Guthrie.

The question will be how long he thought he could hide.

Because the mask fooled the camera. But it did not fool the data. And the data never sleeps.