Leading Through Disruption: Christian Quinn on AI and Police Leadership

The CopDoc Podcast - Season 10 - Episode 171

In this episode of The CopDoc Podcast, Dr. Steve Morreale sits down with Christian Quinn, retired Fairfax County Police executive and founder of Fulcrum Innovations.

Christian traces his path from Massachusetts summer police officer to one of the technology leaders in one of the nation's largest county police departments. The conversation then dives into the rapidly changing world of artificial intelligence in policing.

Steve and Christian explore:

1. How AI is already affecting police operations and investigations.
2. Why agencies need governance, policy, and training before deploying AI tools.
3. The risks of deepfakes, synthetic media, and digital evidence manipulation.
4. How leaders can use AI responsibly to improve efficiency without surrendering critical thinking.
5. What police executives should be doing right now to prepare their organizations for the future.

Key takeaway

This is a candid discussion about the opportunities, blind spots, and ethical challenges that come with AI in public safety. If you lead people, manage policy, or are trying to understand what AI means for policing, this episode offers practical insights from someone who has been on both the operational and technology sides of the profession.

Contact us: copdoc.podcast@gmail.com 

Website: www.copdocpodcast.com

If you'd like to arrange for facilitated training, or consulting, or talk about steps you might take to improve your leadership and help in your quest for promotion, contact Steve at stephen.morreale@gmail.com

Transcript

Intro/Outro 0:00

Welcome to The Cop Doc Podcast. This podcast explores police leadership issues and innovative ideas. The Top Doctor shares thoughts and ideas as he talks with leaders in policing communities, academia, and other government agencies. And now, please join Dr. Steve Morreale and Industry Thought Leaders as they share their insights and experience on The Cop Doc Podcast.

Steve Morreale 0:28

Hey everybody, Steve Morreale back at you coming from Boston. A little chilly up here, but we're headed to Northern Virginia to see and talk to Christian Quinn, who is a retired leader from Fairfax County Police Department and now very involved in AI, which I think you'll find very fascinating. Hello, Christian. So glad to have you. I have been watching. I am a fan of LinkedIn. I'm in LinkedIn all the time and I'm watching what happens, and I saw your name come up a couple of times. I saw actually I I attended one of one of the sessions that you you had. So let's talk about first of all, you're from Massachusetts. You're a product that's now in Northern Virginia. And you went down to Fairfax County. Talk about what made you do that, how long you were in policing, what you did there, and what you're doing now.

Christian Quinn 1:20

So probably a kind of a zigzaggy longer story than we'll we'll tell today. But the short of it is I was at Stonehill College for undergrad in Massachusetts. And I was actually came in as a finance major and an information systems minor, and studying like business and computers. And I took an elective in criminal justice and the professor there, Cape Cod hires summer cops. And I was like, Well, what do they do? And he's like, Well, they put you through an academy and then you go and you patrol like Main Street and the beach, or sometimes you get used like a regular officer, you have full sworn duties, armed. And uh now looking back is like insane to think that you equip a 19 or 20-year-old with a firearm and a handful of training and send them off to go do good. But at the time was working for Coca-Cola in the warehouse. Well, that sounds like a much better job than what I do all summer, and went and did that and really never aspired to go into law enforcement. I have no family members that I have a lot of family members that are military veterans, but none that were first responders in any shape or form. So I went down to the Cape and worked and really got a taste of policing. This is like an awesome job. Great people, and you do get to do good here and there, and got a little taste of that and and and the camaraderie of it and the fun of it as a young man. This is a a neat profession, but it still wasn't what I thought I would do. So I came back to school and I went back for another summer. And as I was, moving through my academics, I thought, well, maybe I would do something with this. But I still never thought like I'd make policing a career. I thought maybe I'd go to work for the federal government. I was thinking I need to get more advancement because this is the 90s where the feds are on a freeze, the crime bill hadn't come out, it was just coming out. So a lot of hiring was stagnant. And as you know, from Massachusetts, it's kind of unique, right? You don't just walk in someplace and apply for a job and you get hired. Civil services, so civil service exam. So I went down to GW to get a master's degree in forensics, thinking, okay, I'll get serious about this as a profession, but I'll kind of, and I was always interested in science. I was always interested in tech. So forensics was kind of a natural progression for me. So I come down to DC and it was great because you get like a real taste of a lot of agencies, a lot going on. And I interned with U.S. Secret Service, I did my capstone, I had access to their forensics lab. I actually was interning in financial crime. So I got there and was doing things related to like Nigerian freak fraud and transnational crime. A really neat group of people, interesting time to be there. I was in the office the morning of the Oklahoma City bombing. That was my first taste of like a tragedy in a in a in a law enforcement setting where I saw so many people worried about their colleagues and stuff. And they ultimately sent me home that day. But I finished my degree and in my program, there was a guy that said, What are you doing after graduation? And I said, I'm applying to some places now and what have you. And I said, those are the days that you take treasury enforcement exam and other things. So he said, Well, we're hiring. And he was a Fairfax County comp. I don't know anything about Fairfax County. It's west of DC in Virginia. It's a big county. We got eight police stations. And he said, you know, we're hiring. So I said, Oh, I'll apply there. And I applied to some other places, U.S. Capitol and some other spots that I was interviewing. And Fairfax just hired me first. So I thought, okay, I'll do this for a few years and then when the Fed thaw comes out. And as it happened, I was in Fairfax County's academy when my Fed process started. So I was in the academy and went down to Quantico to do my PT test. So they started my background. And the short version of this is they wanted me to start their academy, whatever it was, was the week of Halloween, and I was graduating uh Fairfax's academy September 17th that year. Man, I don't have another five months of this in me. I'd like to get some work experience and stay stay in one place for a while. So I ultimately went out and started working for Fairfax and then made it to Detective and Organized Crime in Intel, made it to the jump team quick, doing more high-level operational work and was there when 9-11 happened. And I often tell people that that event being in Intel during that window just changed the whole trajectory and my whole outlook on my career because all of a sudden we were just doing all kinds of work with federal partners, with regional partners that was meaningful and exciting and important. And it was a high-stakes time. And, a year later we're going through the Beltway Sniper event. We had the anthrax attacks in the area. We just had a whole bunch of things going on. And as somebody who had moved for the job, was unattached, no family, I took advantage of all the training and all the things that came along and got involved in a lot of like long-term cases that were exciting and interesting, did a lot of undercover work in more conventional organized crime. And it it was just an awesome time. But I had blinked, and you know, like I think like eight and ten years have gone by at that point. Like owned a house, had met a girl from Northern Virginia and kind of rationalized, this is probably home now, and this is probably the the agency I'm gonna be with. So I'm kind of it it's all it's interesting.

Steve Morreale 6:37

You get it's a tipping point for all of us, right? We're always looking for potentially something better, but that eight to ten year is the tipping point. When I was with DEA, it was it was when I don't know if I'm gonna stay, maybe want to go back to Massachusetts. You know the fools that we are, they always want to go back home. I don't know why. But but then I'd say you time out, then you take a an assessment of, well, where am I at? How many more years do I have to put in? I'm already vested here. I'd have to start over. It's like going to a routine where you are the bottom of the barrel. And so I'm sure that that was one of the deciding factors. You were very successful as you moved along with Fairfax. But I want you to fast forward. You left as a deputy chief, but you also were very involved in cybersecurity, which took you from Intel into cybersecurity and other and other further down the road.

Christian Quinn 7:23

So it was kind of funny because as a as a first line supervisor, I ended up implementing a number of solutions. Like when I was in auto theft, we me and another guy built like a case manager system that we locally hosted that just did us a bunch of favors. When I went to Fugitive, I reorganized how because I had this epiphany one day that I was sitting there, like, you know, the the stress of that job isn't getting up at 3:30 in the morning to go break down doors and be responsible for those outcomes. Because I was a lieutenant when I was there. It was managed. I said to somebody one day, I run a travel agency for fugitives, moving them all around the place and detectives that spend all kinds of money that I got to keep track of. And we gotta legitize this, we gotta automate so much of this. So me and another guy built a system for that. It kind of did a lot of those things. But I think the real big thing that I was entrusted with was when I made captain, I was the commander for World Police and Fire Games. And it's this 10-day event with like 50 some odd venues. And in in the National Capital Region, it was like a bit challenging because you're on the Potomac River, you're in DC, you're in Maryland, you're in Virginia, you're on federal properties, you're all over the place, and just working through all that. And I ended up really adopting a lot, and it's like obviously I did none of this alone, but implementing a lot of technical systems to keep track of things that were pretty new and pretty novel in 2014, even social media monitoring and things like that to do threat assessments and things. And kind of had a zigzaggy career through the executive ranks and had some really fun jobs. Was a was a district commander running a station. That was one of my absolute favorite jobs. And I found my way into policy and directives because the agency was in a funny place. It was that post-Ferguson era. And I like to joke that I don't think there was a single day I liked that job, but that job did me a ton of favors in terms of really learning policy, in terms of really engaging with internal stakeholders and the agency and in the employee groups and just our line level folks all the time, and to look at like issues and then to look at the perspective of community members to deal with special interest groups and to to find common ground and then to craft things that we could write that were operational, that we could hold people accountable on, that made sense. And then I kind of bumped up to like so I had made it to that deputy chief level in Patrol Bureau, and I'm sitting in the office one day, and the assistant chief who I work for said to She's like, Yeah, we're making some promotions and some moves. And she said, We we're thinking about consolidating all of our basically our science and technical resources, and that'll go under one bureau. And I said, Who are you gonna give that to? And she said, Oh, nobody's gonna want that. We'll probably voluntole somebody by way of a promotion and send them to that job. And I was over Patrol Bureau and I said, Give that to me, I'll do that. And I said, Yeah, I mean I have a master's in forensics, I work cases for almost nine years, I'm technically oriented. I said that'd be something for me to build and something for me to work on during the during my the last of my tenure. And at this point, I had processed for a Chiefs role in 2016, was offered the role. Won't get into wear because I had it down, so somebody else is doing that job. But opportunity to come back up to the Northeast, and I ultimately didn't take it, not because I didn't want it, I really did want that job, but from an economic standpoint, it it just made sense. I had a friend and co and colleague, Joe Hill, who was with Fairfax, who was really a mentor to me. He always outranked me by a little bit and I worked for him. But he sent me a text message the last night I was in the hotel on the Chiefs process, and he said, Hey, just so you know, I'm leaving, I'm taking the Chief's job in Ori County, you're gonna be on the short list to get promoted. Just do the math on being a 25-year employee at the next level versus a 20-year employee at your current level. And that math just made sense to to stay for uh another four or five years. So anyway, I landed in that role and I did I had not put in for assistant chief because I really wanted to stay close to the work. Monetarily, it didn't really make that much of a big difference week in and week out or from a retirement standpoint. And I and I wanted to stay closer to the work and build something. So that was a super fun three and a half plus years of being kind of the the tech guy and doing and then having troops come to you and say, like, hey, would you help us build a program? Would you use the juice of rank to help get this over the line? So we built a drone program. I was responsible for procuring body run cameras and implementing those. And we did a mobile crime lab program. We brought ballistic identification to the agency. We just did a a number of things that end of the 21st century, right? Yeah, and it was it was work, but it was fun. It just it felt like and a lot of people don't like that work. A lot of people don't like having to go out and engage stakeholders and deal with budget and stuff like that. But I found it was really meaningful, and especially toward the end of my career, it felt even more important to like help the agency and leave things behind. Troops and line level folks are saying, we need this, we want this, to sort of have a hand in getting some of that done and getting some of that stuff established was was fun. It was it was meaningful to me as well at that point in my career.

Steve Morreale 12:29

So we're talking to Christian Quinn, and he is former deputy chief at Fairfax County, Virginia Police Department and now runs Fulcrum, right? A company that focuses. Uh tell us a little bit about that, Fulcrum Innov Innovation.

Christian Quinn 12:42

So my first stop coming out of the police department was with Brooks Bodden Moore. And Ben Bodden and I kind of met because I answered for my agency's FaceRec program, among other uh programs, but FaceRec being probably the most contentious. And Ben working with Ron Brooks, who's a former chief from Cal DOJ, uh Ben's got more of a true business background and and is a is a veteran of the Hill. And they were working on legislative issues, so we kind of crossed paths and and came to know one another. And I had a conversation with Ben one day about because I was being courted by the private sector, I had my time in. It was the end of that 2020 year, which was a hard year. And I wasn't, I certainly wasn't angry or upset or anything like that, but I was also taking stock of like I've had a a wonderful run in policing. Fairfax is an amazing place to work. It's it's a good agency, it's a well-resourced agency, you've got a supportive community, it's a clean department in that we don't tend to find our way into like the scandals and things like that. There's a high expectation on hiring and training and things like that. It was a great place to work. And and I found that I had, as an as an executive, the f the freedom to to build programs and the and to move the needle and do things. I certainly wasn't in a hurry to get out of there, but I also was looking at I've got my time and and there's opportunities coming in, coming my way. And ultimately, Ben ended up coming back to me and we had a conversation about what do you think about coming to partner with us and work with us? So I did a year with Brooks Bottom Moore, and they were they were fantastic to me. And I I maintain a friendship with them to this day, and and we still collaborate on things, and I'm grateful for that. But I went to them and I said, hey, I'd like to build my own thing. I wanted to be able to do more long-term like boutique consulting to help some startups get off the ground, to do training, to take on policy work, to take on projects that when you're part of a firm that that has a number of clients that you you just don't really have the bandwidth to do those kind of things. So being on my own allowed me to do that, and they were super supportive of me going out and very, very gratefully, thankfully, I got off and running, started working on some legislative issues, started working with some clients. One thing led to another thing, led to another thing, and very and I blinked, and and that's been four years and change, and just having a good run and continuing to evolve, my own offerings and the things that I get involved with and the things that I do for clients and agencies. And I would say my firm is a little unique in that we work with probably an equal number of corporate clients as an and an equal number of gov clients on the policy and the legislative stuff still, but also on helping established companies think about where they're going, connecting them with customers, thinking about how to evangelize their offerings, thinking about policy issues as they develop tech. And then on the training side, that really came about because a single entity reached out to me, hey, would you put together training? And I had done a lot of training during my career. I taught ASCII agent chemical field. I got tapped for could you do a training on this? And ultimately went to instructor school and got some certifications and stuff and then started training. I never wanted to be a full-time trainer. I've also guest lectured at a lot of colleges, and I always joke with the folks that host me who have said, Oh, you should do this full time. I don't want to I get to be the band. I come in, play my set, and leave. I don't have to deal with parents and students and grading papers and all that goes with that. But I do love sharing information and seeing how people receive that and make use of that information. I just don't want to hang out and grade papers and be there every day.

Steve Morreale 16:21

No, I understand that. I'm just saying, Oh no, no, I'm at the end of my career with that. That's my second career, as you well know. But think it's important to make the turn to talk about AI and how you got involved in AI and how all of a sudden I think you are a spokesperson for using AI and policing for any number of reasons. And as I said before, I wrote a book by mistake reading policing with AI. And you and I had a conversation just before that AI is changing as quickly as day-to-day. And and there's so much competition. As I said to you agreed, I have a tendency to try out many, many different platforms: co-pilot, Gemini, Perplexity, Grok, Claude, Chat GPT, and I know you know more than that. And and I'm now a Claude aficionado, but I'm always going to others. What I want us to say and see from you is police chiefs are beginning to pay attention to AI, but they're also nervous. I started as a skeptic. And by the way, we're talking to Christian Quinn. If you remember when we first were introduced to AI, maybe you before me, but you would get this negative, sorry, I can't do that because I'm only trained to 2021. Do you remember those days? And it used to piss me off because I'm thinking, well, what good are you? I'm afraid I need you. But at some point in time, we stopped getting that message and it became fairly valuable in a lot of ways. But for me, when I'm doing training, I remind people you control the machine, not the other way around. You challenge the machine, you say this is not correct, you challenge hallucinations, you expect verifiable sources. So to speak to that, because I know that you've had that same experience, and there's also that fear and reticence, and yet I don't know how a leader can ignore AI and policing today.

Christian Quinn 18:22

A couple of thoughts on this. And I was on a stage Saturday, and and I I closed there with this, and I'm and I've now said this a couple of times recently. OpenAI brought ChatGPT forward in 2022, and my worry is that we have well-intentioned employees who are doing things on their own, whether it's revising police reports, whether it's summarizing interview notes, whether it's like using it in investigations, whether it's using generative to so-called enhance evidentiary images and things like that. And my concern is that somebody's going to make a mess that comes to the forefront and is problematic across policing. And for that particular agency, if they don't have a policy, we're not going to be able to fall back on, well, it's just so new, right? Because in in the public's eye, it's been out for four years. And we've got people out there that are smart folks that they're using it. And now you've got even if you look at the legislative environment, if you look at like what was coming forward in 2024 and 2025 in the way of bills, and there was like 150 some odd bills nationally one year and 200 the following year, those bills, the policy environment was we gotta do something, but nobody knows what to do, which is a terrible policy environment. So you have legislation that's very restrictive or very non-enforceable because the model policies was like aspirational. Trust, transparency, collaboration. Like, well, that's not this is okay, this isn't okay, this is okay for this and not this. That's not operational that you can hold people accountable to. I think the challenge now is a decent policy that people are aware of and adhering to and understand the reasons why they can't do certain things is far better than the perfect policy next year because we we've definitely entered a window where it cannot be ignored. The other thing is we're starting to see bad actors. You had the case in March with Daron Lee using it to establish an alibi. I have personally shown in trainings that it's easy to jailbreak to plan domestic extremist attacks, school shootings, all all sorts of bad things. There have been cases, most of the haven't been adjudicated yet, that it's been used to help plot homicides, building of bombs, and things like that. So this is the new frontier. And then add to that the whole layer of all things with like synthetic media and deepfakes, right? It's digital evidence, authentication was always an issue. I would submit to you the stakes just got drastically higher in the past three years. I was talking to an agency two weeks ago, and one agency gets an IA complaint about one of their officers doing donuts in the parking lot. Well, it ends up not even being a real video of their car. But you've got to work that much harder to disprove the negative. Another agency, an officer is is lunked in coming out of a convenience store where somebody runs up and these kids just were running in and out of your car. They were opening the doors, lock the doors, and they're showing them a video of these kids jumping in and out of this car, and then one of the kids walks by and he's yelling at the kid, putting him on the ground. Just the number of things that are gonna confront us that we we need to both have awareness of in terms of how we make use of the tools, because that tends to be our first, our first thought is like, hey, how are we using this? How are we not using this? And I would challenge those assumptions like, how's it gonna start showing up in investigations? Because it's gonna. And how are we gonna start working through the question of what is real? Because the vast majority of us spent most of our careers. If you had a picture or video something back when I was working a case, not that you had definitive evidence, that led you to believe an investigative theory. Your first thought was on maybe that never happened in the world.

Speaker 1 21:57

No, but it has to be now. Well, that's that's in. Interesting point of view because it strikes me, if you think about you were in the cybersecurity world for a long, long time, and you know damn right well that police agencies go out and they get they get calls and they take and they take reports. You know, sometimes it's as simple as this, you know, hey officer, I was I was duped. There was fraud. I ordered a I I ordered an iPad. And when it got here, first of all, it was it charge it was only for 50 bucks. When it got here, it was an empty shelf. There was nothing to it, but I bought it in China. Well, what does the Fairfax Police Department? What does the Syracuse Police Department do with that? They don't call Shanghai and say, officer, we have a problem. You take a report, and in many cases, you're taking a report for the insurance company, if nothing else. You've seen that, and then you've got the FBI IC3 and all of those kinds of things that say we'll take all of these and then they become overwhelmed and nobody investigates it. Is that gonna happen where we rely on the feds to authenticate or disprove fakes, deep fakes, uh, and they you can't get the answer. So now an agency has to do that, a state has to do that. I'm sure this is what you're thinking of. Who's gonna do that?

Speaker 2 23:08

Here's my biggest worry is that when we talk about the speed at which things are moving. So two years ago, I'm looking at deep fakes through a lens of like, well, the metadata's still there and what have you. And then I'm with a colleague and he showed me in real time how fast you can change metadata. But he also, we we had a good conversation about like provenance, like when an image or a video first comes to exist. So my worry is that you used to be able to look at images, and I go through this in my training, where there's things that contextually don't make sense, that like they don't match the realities of the physical world. And there's things that from a technical standpoint, the pixels do not align in a way that they ought to, where the shadow is like misrepresented or a reflection is wrong, or those kind of things. The problem is that AI is getting so good that there is no means to just look at something and say, this is real, this isn't real.

Speaker 1 24:02

So in an Yeah, the naked eye won't show that most of the time.

Speaker 2 24:06

In a in a realm of where digital forensics, because I had digital forensics under my umbrella, we were we were constantly overwhelmed. Like those people were constantly working hard. And and most days, right, you're triaging. I hate that term, but it most accurately describes what you got to do. And that that residential burglary is getting pushed to the back of the pile because you got child sex abuse material, you got a violent crime, you got a robbery, you got a homicide, you got a stranger sex offense, whatever, right? And this happens at scale across agencies. And I came from a large, well-resourced agency with a lot of people working on digital evidence. And my concern is that especially for small agencies, they're about to get drastically left behind. A drunk driving or a crash or somebody punches somebody in the face or domestic violence, like cri we already have noticed digital has some nexus to nearly every crime, right? From the street robbery to the gang activity, like at the very least, there's a phone involved. And phones are small computers that are connected to the cloud and other portals and apps and things like that now. So a phone isn't just a simple dump it and grab some text messages and contacts and pictures anymore. That's a that that's a true investigation aspect as well. And I think in the age of AI, where you start to have this question around we're talking about what's real and digital evidence in general becomes that much more higher stakes to authenticate, because there's there's a great paper out there that talks about the liar's dividend and what it basically refers to, and it reminds me of like the old CSI effect, where you go to court and defendants now have an advantage because the defense attorney can can raise the question around plausible deniability, and they don't have to like prove that like what you have isn't authentic. They just need to plant the seed that there's a possibility that your evidence isn't real, and now you're going back to the drawing board to prove this is a real image. Well, how do you even do that when you know AI can can mimic metadata, it can mimic the data. Depending on where you recover the image from, you know, it might have been posted online. You haven't you have no way to know where it was before that, where the point of origin or data origin was. Again, you can find some useful information, the metadata, you can find some other things, but I was just gonna say that I I I think I was duped just yesterday.

Speaker 1 26:37

Uh, there's so much stuff on there, and it had something to do with knees, right? Believe it or not. And and this idea, and and I know this is off topic, but but it was Sanjay Gupta having an interview with somebody. It it had CNN at at the top, and as I listened to it, I started to see some some material underneath, some comments saying this is not real. But it sucked me in. It sucked or in because voices can be replicated, uh obviously video could be replicated. The the the the sophistication in which it comes out, a video where where certainly the the lips are are matching the words is just mystifying to me, and I got caught and I got I stopped. But but imagine how many people are going to be duped over and over again.

Speaker 2 27:25

I'll tell you a funny story that just that like just came full circle for me. So last fall, I'm speaking with a woman, and her husband went to the FBI National Academy with me. So she's private sector, and we're having a conversation, and she said, I'm dealing with this guy, and I don't he might not be a real person. So I said, Hey, here's a goofy low-tech, no tech solution to a high-tech problem. I said, When you're on a call with him, and I gave her instructions on like, ask him what his IP address is. And if he says, I don't know, have him stay on camera and give him these instructions, have him open another internet window, type this in, and read you the number. He's not gonna be able to lie fast enough. You'll know that he's making something up because he's gonna be looking at a number and trying to make up a number as he goes. So see what his reaction to that is. The other thing is if he's on camera, is as you've interacted with him previously, have him move his head to the left and the right. And if he's actually doing a synthetic avatar or a mask, most people do not have the source material to do a deep fake of your ears in the back of your head. And what you're gonna find is when you turn, it's gonna pixelate like when you wear the same color shirt or the same color tie as your Zoom background. I forget all about this and go on down the road. So I ran into her in California a couple months ago, and she said, Hey, remember that guy? And I said, Yeah. I told him to do that, and as soon as I told him to turn his head, he dropped from the call, and we've never heard from him again. Well, there you go. And we happened to be at a thing for an industry, you know, cybersecurity AI summit, and we shared that story. Somebody said to me jokingly, like, Great, we're gonna be starting every meeting just with people spinning in their chairs just to make sure everybody's who they say they are.

Speaker 1 29:12

That you're real. You know, so we're talking to uh Christian Quinn, and uh I wanted to ask technology in leadership, how important is it for leaders to now understand the value and the importance of understanding, adopting, and having policies to deal with technology now or and in the future? And that technology can it be everything you were talking about? It could be drones. While you're talking about drones many years ago in Fairfax County, for some people, it's just starting to be rolled out for them. But how important is is it for leaders to understand technology, for the agencies to be sustainable?

Speaker 2 29:54

So two thoughts on this. One is there's an opportunity here for leaders to engage more with their communities, for leaders to realize efficiencies for the the agencies themselves, the people working for the departments and entities that are out there. But here's the here's the disadvantage leaders are at. I think until a few years ago, you could kind of play it off. I I got a person, I got a guy, I got a girl, I got this is my insert, my IT person, my tech person, right? You don't get away with that anymore. Everybody's got to be, you don't need to be a subject matter expert, but you got to be dialed in. You're not gonna really get away with I'm just not very technical, because I think this is suddenly the world that we live in. And our community stakeholders expect that. And our younger workforce, if you just tell them, hey, you're not allowed to use this, you're you're just breeding an environment where you've got all kinds of shadow AI because they're gonna use it. That's like telling them in the early 2000s, don't use the internet, don't use email, right? Although I wish somebody going back in time had told me we don't have to use email. Well-intended people are gonna be looking for efficiencies in their work and ways to just work smarter than harder. So they're gonna use the stuff. The challenge is knowing enough and putting policy in place and also not just not just do this, don't do that, but the why of it, right? Like, here's why we don't want people putting sensitive information into these things. They're not secure at all. And we've surrendered that information and we've seen from the an example with like with OpenAI and the share feature, like inadvertently releasing 500,000 to a million chats into the wild and having those discoverable by Google Search Engine's algorithm. And now they're out there for good. Like, granted, they've shored that up. But where's the next thing? That's not even a data breach. How many police reports are out there in the open on Reddit or some other forum? It's not they're stolen and they're on the dark net. They're just information that's leaked. I heard from an agency recently, like, kind of not not bragging, but like celebrating like this novel idea they had where they take their interviews, they transcribe them, and then they feed all the interviews into ChatGPT to create summaries. I'm not a do this, don't do that guy, but I'm telling you right now, don't do that. Stop doing that today. That is a serious liability. Especially you're risking confessions that come out of that, you're risking like all kinds of like breaches of law enforcement sensitive, like seatis violation. It's just a million reasons to not do that.

Speaker 1 32:22

So let's let me let me stop you for a moment and ask about that, because I think that's important. Uh when I know that firewalls are extremely important. And firewalls can be broken through, you understand that. But for people to understand, agencies to understand, if you're going to do used AI, you're going to put specificity in there, doc documents with names and dates and times and all these law enforcement-sensitive things, they have to be blocked, which means you cannot use an open source.

unknown 32:52

Right.

Speaker 2 32:52

Right. And right. And when you say open source, we're talking like you're paying $20 a month, $200 a year to like ChatGPT or Claude. That's still open. But I think for your listeners, it's really important for people to hear you got to be on an enterprise solution, you got to be on a.gov solution, you got to be on something that the terms of use specify that the vendor cannot access your data, they cannot use it to train their model, and that it is exclusively yours and doesn't leave the network. I recognize that that's limiting, but honestly, that's the only way. And that's why I'm faster to tout this for finding organizational efficiencies than I am investigative advantages. Because you've got to be really well resourced. And I think the cost of this is gonna come down because everybody wants local LLMs, leverage the power of these things. But right now, it's just not super cost effective to gain access to that and have a locally hosted system.

Speaker 1 33:51

So thank you for s for for clarifying that. I think that's so important. And it seems to me that's exactly what should be in part of a policy. But imagine, maybe I'm wrong, but when I'm talking to leaders, I'll say sometimes you're walking out of a meeting and you want to dictate and you want to say something that I just came out of a meeting and we talked about that. I wanted three more cars, but here's the argument against it. Where are some uh uh agencies that have been successful in their arguments for? I don't use an agency name, but the but the problem with that is, and please help me, because I think you know better than I do, Christian, that's I I just dictated something today, and because I am a paid member of Claude, for example, it fascinates me and scares me at the same time that it knows. I I'll give you an example. I'm going up with my nephew and I'm giving him as the first photo, right? And I was saying this is here's my schedule. You know, let me know what my schedule is this week. And it came back and said, This is an amazing milestone for you, especially with you being a sergeant in the army. This is so meaningful. And I think I if I told him that it was about a year ago, but it just is amazing to me. Yeah. So so react to that. You understand, and that's that's the that's the problem that we face by giving up this information.

Speaker 2 35:07

I had a very similar experience, same with Claude, and I'm not knocking Claude, but it was the same experience where quite a while ago I helped my brother with a permit process for his business, which is a food shop in Massachusetts. And it prompted me with a question basically about his business. And I thought that was a chat or two quite a while ago, and it still knows that and it still factors that into are you and same thing. It's a it it is disconcerting. And I like granted, like you have the ability to turn off certain user permissions and things like that. But it's also the the trade-off is like using some of these memory functions, there is utility in it. Because like, you know, you and I are probably doing similar work repetitively. And I do like when it prompts me and says, Do you want this? Hey, you've done this in the past. Do you want me to make something like this again? And there's a functionality to that, but it also hits me as being a bit creepy as well. It is and a little disturbing from an information management standpoint where I forget that that's in there essentially forever until I go back and delete it. Or you know, and in the instance of OpenAI, most people don't realize that they they've been operating for a long time under some kind of injunction in the state of New York that mandates that they maintain every chat. So you can delete your chat and it's not it's no longer in your queue. You don't see it anymore, but it still exists. You could delete your whole account, and every single thing you've ever done on ChatGPT still exists forever.

Speaker 1 36:48

So I I want I want to talk about this. And we talked about it before we started, and that was it it seems to me that uh you know, prompting. Let's talk about prompting and what you what you have come to learn about prompting, whether it is in an enterprise function or open. Prompting becomes important, and I write about that in the book, but it it also uh makes me understand exactly what you just said. AI wants to please, AI wants to continue to work for you. Even when you're done, it's always asking questions. Do you want me to do this? Do you want me to do that? Tell tell me your understanding of that and the pros and cons of that.

Speaker 2 37:22

So what's funny is this is another space that's like changing very quickly right now, and I'll and I'll touch upon that. But early on, I think when OpenAI rolled out ChatGPT, people are like, oh, it doesn't work that good, it doesn't this. And and I think each of us settle into how do we make use of it. So early on for me, a big role task format. This is the role, this is the output, these are the constraints, these are the things I want. This is the tone, this is the format, and just contextually constraining, like I don't want you drifting off these rails. And even when I did like deep research, I'd dedicate I want you to go to these sources, not these sources. I like perplexity early on because it was one of the first ones that gave you responses with citations, and now you can kind of get that out of almost any of them.

Speaker 1 38:08

What about notebook uh what by any chance? Notebook L.

Speaker 2 38:11

I like it more for working on projects and creating outputs. I like the infographics, like the audio briefs. I use notebook alone probably in work, I'll use it to produce things like outputs. Personally, I'll use it to organize my interest in working out and things like that, just because it's like a nice place to go back and look at information and basically query my own sources. But I think recently, when you look at the cheat sheet on Claude and what you can do with that by creating, I am very nervous about, I'm still from a cyber standpoint, I'm not sold on enabling permissions and having bots work locally on my machine. And I recognize I'm forfeiting efficiencies and doing this, but I keep hearing one-off horror stories about I'm gonna clean up your hard drive and it deletes all your personal pictures or it deletes a whole bunch of your emails. You haven't looked at this in a year and a half, you don't need these, right? I don't want to give agentic AI or another any bot such autonomy and authority as to make decisions like that. And maybe it's ignorance on my own part, but I'm not totally sold on the cybersecurity of these. I think there's a guy by the name, I believe his last name is Wilson, wrote an article, and he talks about the lethal trifecta of AI agents. And he talks if they can communicate with the outside world, if they have access to sensitive information, and if they can act like autonomously, it's a problem. And that's kind of how I feel is I I might run something, I'm mulling over like investing in a single machine to run something locally, but they don't want it running on the machine where I actually do all my work. So though that was kind of a long-winded question to the question about prompting, is I think increasingly systems are gonna start adapting to what are the outputs that you like, what are the things you typically ask for. And prompting will become, if you're a super user, less important because the system is already gonna know the types of things that you're looking for. But for the average user getting going on it or messing with it, prompting is still really important. The other thing is there's some research out there that says having a bot make your prompts actually results in a diminished output. It's better to really give a thoughtful prompt. And I've done projects where I do the prompt on Microsoft Word and it's almost a page long. Now, that is probably gonna become an outdated methodology because you're gonna have systems like Claude Cowork that are gonna know what you're looking for based on your past use. I'll do big prompts like that, and then I'll just test and evaluate. I'll dump it in ChatGPT, 5.2, I'll dump it in Claude, and I'll see what's the difference between them. And then sometimes I'll even take an output like that. And I find that these LLMs have like, I don't want to call them personalities because it's like really etomorphizing them, but you know what I'm getting at. There's a quirkiness I find that ChatGPT wants to meet my buddy and pat me on the back and be like, you're so smart, look at how good this is. And Claude will beat me up and be like, this you didn't think. Oh my God. I think it's neat to take like a Chat GPT output and put it in a claude and say, This is what I'm thinking about. What am I missing? What have I not thought of?

Speaker 1 41:22

What are my blind spots?

Speaker 2 41:23

Like it's a cool way to, it's definitely more work, but it if you want to do a thorough job on something and really use these things as a brainstorming tool. And you can, yeah, like I we mentioned in Chat GPT and Claude a lot, you could do it with Gemini, you could do it with Grok. I mean, each one of these things has their pros and cons. If I were doing something that I wanted a baseline real-world sentiment, I'd probably reach for Grok because it's tethered to the Twitter verse and you get like different outputs. If I'm doing something that's serious for business or a policy thing, that's not the tool I'm going to reach for because I'm worried about all the things that bleed over that and there are less filtered responses that can result in bias and problems like that. But anyway, the point being, I think the utilization of multiple tools and still utilizing longer prompts from a brainstorming standpoint, at least for me anyway, that gives me more to work with. And it, if nothing else, it gives me ideas that I couldn't possibly think of all these things independently on my own. There's almost always some light bulb moment that I need to think about that a little bit more. And, you know, like you mentioned, blind spots. I think all of us do that when we're writing, when we're researching, there's just something that never even popped into your head. And you can there's only so many folks you can bother in their day and be like, hey, what am I not thinking of? And this is kind of like a collaborative partner. Yeah, that's the way it's not a person.

Speaker 1 42:41

Actually, that and that's when I wrote the book, I said it's your thought partner, that you can't you can't do it yourself. I I think the fear that we have, and obviously it's different in policing than it is in academia. I allow students to use it under supervision. Yeah. By the way, they have to attribute it. And then I attribute whenever I write something, I'm saying this was generated by myself, my ideas that were enhanced by a particular LLM. But I I remember arguing, so we're talking, Christian, here we go now. Uh and I hope the audience is is understanding the conversation that that I'm intending to have is that AI can work for you, but you have to be careful. I think that's what you're saying, too. You have to be very careful as it as it relates to the policing world. Can it help you potentially write policy? Over and over again, what I'll do, Christian, is I'll say to it, I'm looking for a policy on body worn camps. Please go out and find policies that have been in effect and assess them for the beginning and the end. In other words, policies are made to be changed. Not they're not when you first wrote your policy on BWCs, I'm sure you realize there's some unintended consequences. We better fix this before we're held accountable. So it's it's an ever evolving policy, it's an ever evolving process.

Speaker 2 43:57

And I think for AI, that becomes particularly important. When you adopt something that from a system standpoint, simply making use of it, not always, but just having user inputs or vendor updates can cause a different type of drift. And that's why I'm a big fan of thinking about anything that's reliant upon AI as being an iterative process, as being something that like you're going to keep an eye on those guardrails. And unlike other things that you adopt, it's not set it, forget it. It's it's deploy it, keep an eye on it, would be would be my counsel.

Speaker 1 44:33

I think you're you're right. As I had said over and over. Well, I'll give you an example. I I'm writing a book and I I went back to it and said, hey, here it is. I'd like you to act as a developmental editor and tell me where tell me where I've made some mistakes, tell me where I'm drifting, tell me where tone and all of that kind of stuff. And it started, like you said, it was Claude and it beat me up. And finally I said, I remember seeing an article that said, challenge AI. And when you challenge and push back on AI, you will get better output. Don't accept it at face value. And so I said, This is my book, these are my ideas, this is my experience with DEA. It's a novel. And I and I don't appreciate you taking it over. And it's just completely correct. It's you're right. This is it was one of those moments where you have to push. You cannot accept thing at face value. And what I started to say was this that I know that so many people who are faculty members are worried about students using AI as their mind. In other words, using critical reason. And I think that's the big mistake. And that to me is extremely important. And I'd love your take on pushing people who are using it and how to use it better, and how to use the challenge to get better output and not accept it on face value.

Speaker 2 45:48

So I think the synergy between a person and a machine is this the person gets going on it based on their expertise, their perspective, and their original idea. And then they enhance that or check that or do whatever using the machine to get more out of that or to perhaps clean it up a bit. But I think the problem with AI is the temptation to outsource the thinking to it. And you'd be like, oh, I got, and I gave it a, I did a LinkedIn post on this because I was writing an article and it was on something technical. And it was actually on a topic. I'd already written an article on the same topic. So I thought, well, I'm going to use some deep research, uh, get the new data, I'm going to feed it my old article, I'm going to tell it, like, write a similar article and just paraphrase it. And you know, the whole process didn't take me 30 minutes. But I ultimately read it and I was like, I don't like this, and scrapped the whole thing and took the research, took out a pad of paper, looked at my old article, thought about like how I think about that issue differently, and just started writing out an outline. Then I started framing out some thoughts, like just old school, like pen in hand, tactile piece of paper, because I like writing like that and organizing my thoughts in writing. As technical as I am, like I keep a paper to-do list just because I like this, I like the satisfaction of like adjusting with the city. Crossing stuff out, right? And crossing stuff out that I got done. Yep. Yeah. And I think that is our challenge, is especially for like younger learners, younger workers, is we have this fallacy of expertise. And there's like now good research out there that shows, and I I do want to come back and correlate this to policing because I think it's really important as we think about what we're adopting. And if you looked at like that MIT research last year, where they actually put control groups together, they had one group write an essay using an LLM, another group using Google, and another that only could write it out of their head. And then they actually didn't just survey them, they put EEGs on their head to look at like what parts of their brain were lighting up, how bright they're lighting up. You know what I mean? Like the distributed networks, how active are they? And probably no surprise, the people that wrote the essays had the greatest activity in their minds afterward when they were asked about the topic. And the people that use Google had a more diminished, and the people that use ChatGPT really showed minimal. And then not only that, but when they asked them to identify their own work and quotes from their work, the people that use ChatGPT, some couldn't even identify their essays. So I bring this back to there's a forfeiture there, and there's a cognitive debt in using these tools. And I often think, and I'm not knocking automated reports, if we're going to use automated reports as a profession, we need to seriously think about how we train our new workforce and our existing workforce to document things. Because the the when you get busy, the ease of say of looking at the first draft that an automated report writer gives you looks good. Click. And I think the exercise of looking at your notes, of thinking about what you experienced when you got there and not just what people said, because that's one thing. I mean, the other thing is like I remember being like a midnight officer with about 18 months on and having this epiphany all day, every day, everybody lies to me. Because I'm not knocking humanity. I'm just saying, like, you get you get to places and people's default is probably to not tell you the truth. So a lot of times what's verbally said really is secondary to what you observe, right? Like odors, signs of a struggle, damage, like, you know, like I we we could probably swap up a million anecdotes on this. I went on a suspicious event that ended up being a rape that was like a terrible case that like nobody in the entire equation told me what actually happened. And it really came down to confronting people on how did this door frame get broken? Who just left here? All these things that are just good police work that if you default to the first, you're gonna miss that. So I think all of these things that are contextual, that are non-verbal, we need to really emphasize to our workforce, be documenting that. Nobody's documenting that except for you, what you see, what you smell, what you hear, what you perceive. Just somebody sweating and even violent encounters. I often say, like, the worst physical assaults I was ever in with suspects, they didn't begin with them announcing, and now I'm about to assault you. It's like nonverbal, it's a clenched fist. It's uh and then take that to the extreme where you have like your your sociopath who's been in in and out of facilities their whole life, predatory mindset, they pounce on you, and there's no indication that it's coming. So I think the point is this if we're gonna use tools like automated report writers, we need to really rethink how we train our workforce to like capture, because this the goal is not to have the most beautiful grammatic formatted prose. The goal is to have a representation in the most accurate sense of what happened for either a trier of the fact later to make determinations and to ask you about for you to refresh your recollection or for a detective working that case to read it at first blush and say, okay, this is the direction I need to go now with the information that was available at the time for the first person that went out there. And I and I and again, I'm not knocking automated reports, but I I have a lot of concerns around both like people's outsourcing of their thinking in a situation where you've you've got to be on your toes and really mindful of the fact that most of the things that are gonna be represented verbally are probably gonna be, at the very least, inaccurate because people are excited and emotionally upset and things, or they're just gonna straight up lie to you because there's a self-preservation agenda there where they don't want to admit wrongdoing or their culpability or what they did.

Speaker 1 51:31

So you know, as I as I'm thinking, and we have to wind down in terms of time, but when we think about body warm cameras and what we've been talking about about deep fakes that I I fear down the road that the video doesn't lie, and yet that's what we would like to think, but there are gonna be plenty of people who say, oh yeah, sure, Christian. Yeah, you you weren't the first to hit him. This this has been doctored, right? Right. It's that it's that same challenge on the other, on the other hand, that that it's gonna backfire. But let me let me wind down by asking a couple of questions. As in your work, what are you leaning towards? What's the concerns that you're hearing from vendors, but more importantly from police agencies to try to figure out how to get engaged, get involved in AI the right way?

Speaker 2 52:23

So here's the thing: it's gonna be like everything else. It is a tool. It's not magic, it's not sentient, it doesn't know anything. We can wield this and absolutely realize efficiencies. We can do a lot of good things with it, it can make our lives easier. It is imperfect, and it's like a lot of tools. And I think we have to master where are those imperfections and how do we account for inaccuracies, bias, just hallucinations and other shortcomings of the tools. And I think from a vendor standpoint, we don't want to over-engineer everything. AI has become this sexy buzzword that's being attached to everything. I don't think that's necessarily needed. I think when you look at like some of the better risk frameworks out there, you know, like at like at MIT, one of the brightest minds was like, you know, the first prong is like, do we need AI for this? Like, what's wrong with the way that we're doing it? And there's a there's a simplicity to that that's like really eloquent where it's like, you know, that's a good point before we start like just bolting AI onto everything. And then I think the other piece is this is going, there's there's gonna be an element of the hype that settles down where this is like more widely understood and we start moving into a place of like, okay, and I think we're moving there now, right? When you think of some of like the Claude capability right now, it's pretty remarkable, right? And I'm I'm excited to learn more about this next evolution because in some of the trainings I've done, I've seen stakeholders who are not terribly technical build what's basically like a working prototype on an app for like for wellness, applicant tracking. So the barrier between an idea and a product just got lower. But we need to like make use of these things responsibly. And and you know, 15 years from now, this is just gonna be ubiquitous, right? But there's still gonna be leaps and jumps in terms of capability and advantage. But right now we're kind of in the the thick of like the the Wild West part of it. And figuring out how we're gonna wrangle that in, how we're gonna make use of it responsibly and ethically, is probably our challenge going forward right now. And looking at things like agentic AI that's got so much promise, but also like a lot of vulnerability and a lot of opportunity for drift and things to go off the rails, we need to be mindful of that as well.

Speaker 1 54:44

Well, we've been talking to Christian Quinn, and he's in Northern Virginia today. And as you heard at the beginning, a former deputy chief with Fairfax County and now with Fulcrum Innovations and very active in speaking his mind and helping both vendors, uh, corporations and police departments figure out this uh new world of AI. What question didn't I ask, or what area did I miss that you think is important for the audience to hear about AI?

Speaker 3 55:20

Oh, that's a good one.

Speaker 2 55:22

How about uh for I think for policy considerations? This is this is like my like what's probably the most important thing to think about policy, right? I would say this. Don't have a policy governing all AI because the we're then you're boiling the ocean. I think the need right now is a policy around generative AI. When can they use it? What's it okay for? What solutions are available, what training are you gonna give them? What communication are you gonna put out externally, internally, with respect to utilization? And then anything that when with more tangential, like algorithmically enabled processes and things like computer vision and those, like all those kind of things that are a form of perhaps some type of AI, those all get their own policies. Like they all, whether you're using predictive analytics or facial recognition or some kind of like data integrator, those all need to have their own governance. And I think this is the thing right now. Like, what is the most important thing right now? It isn't the whiz-bang factor, it's not the look what we can do with it. It's how are we gonna put meaningful governance around this? Because if we don't do that ourselves in this profession, somebody's gonna do it for us. And it's probably gonna take the form of legislation that's not terribly favorable and not terribly helpful, perhaps well-intended, but more restrictive than what's needed. But that will be the default if we don't manage ourselves in this kind of emerging space.

unknown 57:00

Great.

Speaker 1 57:01

Well, thank you. It's been a pleasure to meet you and talk with you.

Speaker 2 57:05

It's been really a joy to chat with you.

Speaker 1 57:07

Well, it's been all over the place, but I think AI is so important and it has a different perspective when we when you start dealing with law enforcement-sensitive things. But I think it can be very uh valuable as you and I have have said. So thanks so much. We've been talking to Christian Quinn, and I appreciate your input and your innovative thoughts on this. So that's another another another episode of the podcast is in the book. You've been listening to Steve Morelli's Cop Talk Podcast. We've been with Christian Quinn. Thanks for listening. Uh, it amazes me that now I'm seeing people from 137 countries listening. Thank you for that. And for those of you who are out of the country, give me a give me a a buzz. Let me know if there's somebody in your organization I should be talking to because you feel that they are a leader, that it's innovative. Take care of your people. Thanks very much. Have a good day.

Speaker 58:01

Thanks for listening to the Cop Doc Podcast with Dr. Steve Moriali. Steve is a retired law enforcement practitioner and manager, turned academic and scholar from Webcester State University. Please tune in to the Cop Doc Podcast for regular episodes of interviews with thought leaders and police day.