Report Says AI That Hallucinated A Cop Into A Frog Is Making Utah Streets ‘Safer’

from the that’s-not-AI,-that’s-LSD dept

AI can be useful. But so many people seem to feel it’s nothing more than an unpaid intern you can lean on to do all the work you don’t feel like doing yourself. (And the less said about its misuse to generate a webful of slop, the better.)

Like everyone everywhere, police departments are starting to rely on AI to do some of the menial work cops don’t like doing themselves. And it’s definitely going poorly. More than a year ago, it was already apparent that law enforcement agencies were just pressing the “easy” button, rather than utilizing it wisely to work smarter and faster.

Axon — the manufacturer of Taser and a line of now-ubiquitous body cameras — has pushed hard for AI adoption. Even it knows AI use can swiftly become problematic if it’s not properly backstopped by humans. But the humans it sells its products too don’t seem to care for anything other than its ability to churn out paperwork with as little human involvement as possible.

The report notes that Draft One includes a feature that can intentionally insert silly sentences into AI-produced drafts as a test to ensure officers are thoroughly reviewing and revising the drafts. However, Axon’s CEO mentioned in a video about Draft One that most agencies are choosing not to enable this feature.

Yep. They just don’t care. If it means cases get tossed because sworn statements have been AI auto-penned, so be it. If someone ends up falsely accused of a crime or falsely arrested because of something AI whipped up, that’s just the way it goes. And if it adds a layer of plausible deniability between an officer and their illegal actions, even better.

Not only is the tech apparently not saving anyone much time, it’s also being abused by law enforcement officers to justify their actions after the fact. But it’s shiny and new and seems sleek and futuristic, so of course reporters will occasionally decide to do law enforcement’s PR work for it by presenting incredibly fallible tech as the 8th wonder of the police world.

Sometimes reporters bury the lede. And sometimes their editors decide the lede should be buried by the end of the headline. That appears to be the case here, where Mya Constantino’s reporting isn’t exactly what’s being touted in this article’s original headline.

As can be observed from viewing the URL, the current headline (updated January 1st) wasn’t the original headline. The Wayback Machine tells the real story. This article was originally published on December 19, 2025 with this headline:

That headline (which reads “How Utah police departments are using AI to keep streets safer”) was immediately followed by these paragraphs:

Here’s a direct quote of those leading paragraphs:

HEBER CITY, Utah — An artificial intelligence that writes police reports had some explaining to do earlier this month after it claimed a Heber City officer had shape-shifted into a frog.

However, the truth behind that so-called magical transformation is simple.

“The body cam software and the AI report writing software picked up on the movie that was playing in the background, which happened to be ‘The Princess and the Frog,’” Sgt. Keel told FOX 13 News. “That’s when we learned the importance of correcting these AI-generated reports.”

Fortunately, those paragraphs still remain in the updated post, which now contains a headline that makes a lot more sense:

The headline (accompanied by a short video of a tree frog) says:

Ribbit ribbit! Artificial Intelligence programs used by Heber City police claim officer turned into a frog

While I can understand why a small news outlet (albeit one that’s a Fox affiliate) might decide to play nice with the local cops rather than call out their software failure in the headline, it really doesn’t make it acceptable. My guess is the original headline was about maintaining access to officers and officials. At some point, someone realized the stuff detailed in the first paragraphs would probably attract more attention than some dry recitation of cop AI talking points.

But even the belated headline change doesn’t really make anything better here. There’s not really anything in the article that demonstrates how AI is making anyone safer. The article also notes that two different AI programs are currently being tested (Code Four, developed by a couple of 19-year-old former MIT students) and Draft One, which is part of Axon’s vertical integration strategy. That was the product that turned a cop into a frog, which probably explains why the reporter’s ridealong (so to speak…) only involved use of Code Four’s AI.

The reporter was on hand for a faux traffic stop that was later summarized by the AI to (apparently) demonstrate its usefulness. The journalist points out that the AI-generated report needed corrections, but at least didn’t turn any of the participants into a Disney-inspired character.

That being said, there’s nothing here that indicates these products will make streets “safer.” Here is the entirety of what was said about the tech’s positives by Sgt. Rick Keel of the Heber City PD:

Keel says one of the major draws is that the software saves them time, as writing reports typically takes 1-2 hours.

“I’m saving myself about 6-8 hours weekly now,” Keel said. “I’m not the most tech-savvy person, so it’s very user-friendly.”

Giving cops more free time doesn’t make streets safer. It just means they have more time on their hands. That’s not always a good thing. Of all the things that need to be fixed in terms of US policing, writing reports is pretty far down the list. It’s what’s being done with this extra time that actually matters. Pursuing efficiency for its own sake makes no sense in the context of law enforcement. The statements by this PD official raise questions that were never asked by the reporter, like the most important one: what is being done with this saved time? And if something still requires a lot of human activity to keep it from generating nonsense, is it really any better than the system it’s replacing?

One thing is for sure: AI doing the menial work of filing police reports is never going to make anyone safer. On the contrary, it’s only going to increase the chance that someone’s rights will be violated. And because law enforcement agencies refuse to be honest about the risks this poses and the fact that it appears only officers who don’t like writing paperwork will benefit from this added expense, they shouldn’t be trusted with tech that will ultimately only make the bad parts of US policing even worse.

Filed Under: ai, failure, hallucinations, heber city police, police accountability, rick keel, utah

Companies: axon, code four