Here's how one Twin Cities police department is using AI for police reports

NINA MOINI: Artificial intelligence is changing how many people work. In several Twin Cities area police departments, it's taking on some of the paperwork that comes with law enforcement by helping write police reports. Axon, the company behind TASER, has an AI tool that helps officers write reports using body camera footage. The tool is called Draft One.

And as KSTP TV first reported, three departments in the metro area are using it, but some prosecutors and civil rights advocates have raised concerns. Spokespeople for Bloomington and Brooklyn Park Police said a group of officers within their departments are piloting the tool. And joining me now is Lieutenant Nate Tennyson from the Eagan Police Department, which has been using draft one since June. Welcome to the program, Lieutenant.

NATE TENNYSON: Good afternoon, Nina. How are you?

NINA MOINI: I'm doing well. Thank you so much for your time. I wanted to start, if you would, by just walking me through how Draft One is being used in your department, and how is it going?

NATE TENNYSON: So for Eagan, it is a tool for officers just to give a little background. Officers can simply, if they want, still type their own reports. We also have been doing transcription for years or longer than I've been here. So previous to 1998, we have record staff that we can dictate our reports, and there's record staff that will transcribe those for us and get them back to us for review, but that usually takes at least a day.

What Draft One does is it uses objective AI that only uses our audio from our body worn cameras for us talking to a person, receiving information about what we may be there for, and then it is just transcribing that in a quick, immediate form where it's done within five minutes so the officers can see exactly what is said and then make edits to it. So it can make the process extremely efficient if the officers choose to use that type of method of creating a report.

NINA MOINI: Do you feel like it is saving time? I'm wondering what types of crimes are involved or what types of reports, are they more low level type offenses?

NATE TENNYSON: They are. And that's a great question because as of now, Axon Draft One has a parameter of only having a maximum number of words that it can do. So on a larger, more complex case, if this product was used, it would be more trouble than it was worth because it only can take 850 words. So it would condense some information.

It wouldn't be wrong, it would just condense it and prioritize different information. So it's probably right now more of the smaller cases until they open that up. And it also is easier to organize when you have a lot of intricate pieces from say, a domestic.

There's a lot involved in a domestic case where there could be information that we're seeking that from previous assaults that may have happened or previous forms of domestic violence that's happened to these people and sometimes it does categorize those and arrange them in the way that it should be in a report.

So most officers are using this for the smaller ones. Maybe that your car was broken into and there's some pretty quick facts that come and you and I discussing it. If I use Draft One, it's basically everything that you said and everything all the questions I asked you will just be transcribed into a report form.

NINA MOINI: I think some of the concern, like from the American Civil Liberties Union and others, is that this might just be just the beginning. We're talking right now about lower level offenses, but I think in many industries, people are thinking about how is AI going to be useful, and what about if it's not deemed to be objective in the way that it's constructed? And what about if it's constructing a narrative and an officer reads it and it's not exactly what happened?

Well, now, has it implanted a different seed or a different idea or a different narrative in the officer's mind? Some of these concerns, I wonder what you think about them as it pertains to the future and if AI starts to become an even greater process, which it seems it's headed that way as it relates to drafting reports and just police work in general.

NATE TENNYSON: I think the one thing that we have to remember is that some of those concerns are coming from what everyone knows is the generative AI from consumer products. What's different about this is it's objective AI that only uses our body cameras, audio. So there is nothing to do with it making any assumptions about what happened in a video, it is only about our audio. So if I was using my body camera now for this conversation, the only thing that would come out is only what you and I spoke about.

And then the most important piece to remember on that is after the safeguards that are in place that the officers have to check every paragraph, they have to actually physically remove data that's in that paragraph in order to produce this report. And they can make edits to make sure that there isn't something in there that was inaccurate because we were in a louder environment.

And then what we also choose to do at the Eagan Police Department is, Axon, we can allow it to put in, I guess, I would call it wizards and goblins statements, and they're just outlandish statements that would go in the report randomly throughout the report to ensure that our officers are going through checking and reviewing it and then putting their stamp of approval on it that they are attesting that they have gone through this. And this is an accurate report for what call that they had just responded to.

NINA MOINI: So yeah, you're saying it's not like nobody ever looks at it again, they go over it diligently, is the idea to make sure that it all tracks with what happened. It does sound that right now, like some prosecutors at the county level, like the Dakota County attorney's office, are not accepting reports written by AI. Obviously, Eagan in Dakota County, I guess the city attorney is accepting them. What do you think about that process? Is it hard to use it if the county isn't accepting it?

NATE TENNYSON: We're getting there. With any new technology, groups can decide a little slower. The Dakota County attorney's office is great. They're working with us and we're getting very close to a resolution, whether that's just piloting it with them first or whether they're going to just allow us to fully start using it.

The main thing that I would hope that people understand is that our reports are really just notes, they're not evidentiary. The only evidence is us testifying in court to what we saw, what we observed. So police reports are not an evidentiary piece that is a court document until the courts deem that we've verified that the information there is accurate and correct.

NINA MOINI: Do you sit with other leaders, maybe in the department or other law enforcement leaders, and talk about just the use of AI and creating best practices and what the future holds because, again, this seems like just the beginning of how AI could be integrated?

NATE TENNYSON: I agree with you. And yes, we continually are meeting with our counterparts, meeting with attorneys and trying to look on the best practices that we can use. This one is, again, like I said, it's different because it's very objective AI that's only based on audio. So it's not making any assumptions of anything it sees or any parameters, it is physically just almost like a transcription tool, is probably the easiest way to describe it.

NINA MOINI: Will defendants know if a police report being used in their case was created by AI?

NATE TENNYSON: Or assisted to be created because you have to remember that the report might be created, but then I may make edits to it because I want to add more detailed information. But yes, by our policy, along with, I think, any other agency using it, there's a default footer that goes in there. We just are using the default one that came from Axon. We didn't need to change anything else. So that immediately just tells you that this was used with the assistance of AI to create this report.

NINA MOINI: And just lastly, Lieutenant, is there any other area where you're exploring AI or where you think that it could be helpful? Because again, I think if you talk to many members of law enforcement, they'll say that the report process and that process is really time consuming. Are there other time saving measures that you're looking at as a department that would include AI?

NATE TENNYSON: Yep. We've already instrumented another Axon tool called Translator that uses our body cameras that acts as a translator for different languages, and it's certified up to 57 languages. So it has become a huge help for us so that we can be relatively hands free, and that when someone speaks a different language, that we can at least communicate with them, as opposed to maybe going to a Google translator.

This one is more secured and verified. It is not language translator that we can use for any court. Or if say, we are taking a statement from them, we would then go back to systems that are already in place that have been vetted, that we can use. It would be a human translator through things such as language line. But for conversational piece on a traffic stop, when someone needs help, needs directions, whatever it might be, it's been an invaluable tool so far since using that to be able to just communicate with someone else and get them answers that they need.

NINA MOINI: Lieutenant Tennyson, thank you for coming by and talking about just how to use AI. It's important, and it's something we're all going to be facing as a society. So I appreciate your time. Thank you.