Four Black high school students in Baltimore were ordered to their knees at gunpoint by police last year, after their school’s AI gun-detection software mistook a bag of Cool Ranch Doritos for a firearm. As Ars Technica reports, the teens are now suing the school district, police, and the company behind the software. The case raises a pointed question: what happens when officials trust an AI alert more than their own judgment?
A canceled alert that still sent in the police
It was October 20, 2025. Taki Allen, Yui-Osa Ogbeide, Devon Thompson Jr., and Keith Winfield III were winding down after football practice, sharing snacks and waiting for rides. The school’s AI gun-detection system, made by a company called Omnilert, flagged Allen holding the chip bag as a potential gunman.
Within two minutes, school officials reviewing the camera footage recognized the Doritos bag and canceled the alert. They notified police that no response was needed. That should have been the end of it.
It wasn’t. The school’s principal, Kate Smith, apparently never received proper training on the Omnilert system. She spotted the alert in a text message, missed the cancellation, and texted a school resource officer to deploy police anyway. Her own message acknowledged the object was “maybe not” a gun.
At least six patrol cars arrived with lights flashing. Some officers came out with guns drawn. The students were separated, ordered to kneel on concrete, handcuffed, and searched. Nobody told them why. Only after police found no gun did the teens learn that a chip bag had set the whole thing in motion. One officer allegedly muttered “stupid AI cameras” and “fucking Omnilerts” under his breath.
Why the system kept failing
Omnilert is an AI visual gun-detection tool, meaning it scans camera footage automatically and flags anything that looks like a firearm. The company describes it as designed to flag anything that “could be a firearm” for a trained human to review, because missing a real gun is the worst possible outcome.
The problem is that “trained human review” part. According to the lawsuit, more than 99 percent of Omnilert alerts across Baltimore County Public Schools were false positives, meaning they flagged something that turned out not to be a gun at all. That figure comes from a report by the Office of the Inspector General (OIG) for Education, which investigated the incident.
Think of it like a smoke alarm that goes off constantly when you make toast. Eventually, people start ignoring the alerts, or in this case, acting on them without checking whether the fire is real. Smith had never been trained on the system and acted on an unverified alert she only partially read.
What this means for students in AI-monitored schools
The four teens are still living with the fallout. They report anxiety, fear, and lasting emotional distress. Allen now pays for private car services rather than waiting outside school for a ride. Winfield calls his mother multiple times after practice just to confirm she is coming. All four say they feel less safe on campus and less trusting of police.
Their lawsuit asks a court to bar Baltimore County Public Schools from using Omnilert until proper training protocols are in place. It also claims their constitutional rights were violated through unlawful search and detention. Damon Hewitt, president of the Lawyers’ Committee for Civil Rights Under Law, which is representing the teens, put it plainly:
“Overreliance on artificial intelligence technology mixed with inept communication and coordination among key officials is a toxic mix.”
BCPS declined to comment beyond a statement promising counseling for students. Omnilert said it is reviewing the complaint and expressed regret for what the students experienced.
What to watch
The court’s decision on the injunction will signal whether judges are willing to pause AI safety tools when false-positive rates are this high. Watch also for whether other school districts using Omnilert, which BCPS helped make the largest deployment of AI visual gun detection, revisit their own alert protocols. The OIG’s recommendation for twice-yearly staff training is a low bar. Whether schools actually clear it is another question.