The Officer and the Frog
A real story from Utah that explains everything you need to know about AI — and why this course exists.
Start Here
You don’t need a computer science degree to understand AI.
You need one story. And it involves a frog.
Heber City, Utah — 2024
A patrol officer in Heber City, UT responded to a disturbance call. Like always, his body-worn camera recorded everything — the shouting, the radio noise, the chaos of the scene.
Afterward, he used an AI tool called Code Four to draft his incident report. The tool listens to the body-cam audio and writes the narrative for you. It does in under a minute what used to take two hours.
The draft came back. It was mostly right.
Except for one detail: the AI had picked up background noise from a television nearby playing The Princess and the Frog. It processed that audio as part of the incident.
The AI wrote that the officer had shape-shifted into a green bullfrog.
Why This Story Matters to You
Read that again slowly — because this isn’t a joke about bad software.
An officer was about to sign a legal document containing a fact the camera never recorded. If that report had been filed and later surfaced in court:
- It’s Brady material — evidence favorable to the defense that must be disclosed
- It’s grounds for a wrongful charge getting thrown out
- It destroys the officer’s credibility on the stand for years
One signature. Career risk. Because an AI misheard a cartoon.
What Is a Hallucination?
In AI, a hallucination is when the system confidently generates something that isn’t true.
The AI doesn’t know it’s wrong. It doesn’t flag uncertainty. It just produces text that sounds plausible — and in a legal document, that’s exactly the problem.
This isn’t a fringe bug. It’s a built-in characteristic of how these tools work. Every AI report-writing tool on the market today has this risk.
The question isn’t whether hallucinations happen. They do. The question is: do you know how to catch them before you sign?
The Promise of This Course
This course won’t tell you AI is good or bad. It’s a tool — like any other tool in your kit.
What we will give you:
- A clear picture of what AI tools are already being used in departments like yours
- A straight read on the real risks — not vendor marketing, not union fear
- A practical framework for evaluating any AI tool before you trust it with your name
- The specific compliance rules that govern what AI can and can’t touch in your work
By the end, you’ll be able to catch a hallucination before you sign it, identify a policy violation, and evaluate a vendor’s pitch with clear eyes.
Read this AI-generated incident report excerpt and identify what’s wrong:
“At approximately 22:14 hours, Officer J. Torres responded to a disturbance at 412 Bell St. Four individuals were involved in a verbal altercation. Officer Torres, who briefly transformed into an amphibian, observed the subjects disperse upon her arrival. No weapons were recovered. Case closed.”
What did the AI hallucinate?
Write down your answer before reading further.
The hallucination: “Officer Torres, who briefly transformed into an amphibian” — a fabricated detail that has no basis in the body-cam footage or any witness account.
Why it’s dangerous: An officer who signs this report has legally attested to a document containing an invented fact. Even if obviously absurd, it creates a chain-of-custody question about the integrity of the entire report.
The lesson: AI hallucinations aren’t always absurd. Most are subtle — a wrong address, a misattributed statement, a detail from background noise. The frog story is memorable because it’s extreme. The real threat is the 20% you almost miss.
80% accuracy sounds reassuring. It isn’t.
If you process 10 reports a shift, an 80% accuracy rate means 2 reports per shift contain errors. Across a year, that’s hundreds of legal documents with potential hallucinations bearing your signature.
The human-in-the-loop requirement isn’t a suggestion — it’s your legal shield. Treat every AI draft like a rookie partner’s first write-up: trust but verify, line by line.
An officer files an AI-drafted report without fully reading it. The report contains a hallucinated detail. Who is legally responsible?
Reveal Answer & Explanation
The officer is responsible — not the software vendor, not the department, not the AI.
When you sign an incident report, you are legally attesting to its accuracy. The AI is a drafting tool. It has no legal standing. You are the human-in-the-loop, and that role carries full accountability.
This is why the Washington State AI Task Force is pushing for mandatory officer attestation of accuracy — a formal, named requirement that the officer has verified the content before signing. It’s coming. Get ahead of it now.
Up Next
In Lesson 2, we’ll look at the specific tools already being purchased by departments across the Pacific Northwest — and what they actually do.
The frog is just the beginning. 🐸