Finnect AI Insights

AI Business · · 6 min read

When AI Sees a Gun in a Doritos Bag, What Really Went Wrong?

We dig into the 'technological and human errors' behind an AI security alert gone wrong, and what it means for your business.

Doritos

Last fall, on the evening of October 20, four high school students in Maryland had just finished football practice. Their coach handed out bags of Doritos. Taki Allen, a junior, grabbed a few and waited for a ride with friends near picnic tables.

What happened next wasn't a ride home. Six local police cruisers, lights flashing, suddenly surrounded them. Many officers unholstered their weapons, ordering the teenagers to their knees, handcuffing and searching them. The students were left "fearing for their lives".

The reason for this terrifying scene? An artificial intelligence (AI) gun-detection system, Omnilert Gun Detect, had flagged Allen's bag of Cool Ranch Doritos as a firearm with 90% confidence, according to Google Search research notes.

The students are now suing Omnilert, the school, and the police, alleging unlawful detention and search due to "technological and human errors." A false AI alert is one thing, but how did a bag of chips turn into a full-blown police incident? That raises three questions.

Question 1

What Exactly Happened with Omnilert's AI System and the Doritos Bag?

The incident unfolded on October 20, 2025, after football practice at a Maryland high school. Taki Allen, along with Yui-Osa Ogbeide, Devon Thompson Jr., and Keith Winfield III, were waiting for a ride, enjoying Doritos bags handed out by their coach.

Omnilert's AI Gun Detect system, designed to scan security footage for firearms, flagged Allen's bag of Cool Ranch Doritos as a potential weapon. The system reported a 90% confidence level in its detection, according to the lawsuit reported by Ars Technica.

For context, AI gun-detection systems use computer vision to analyse live camera feeds for shapes, movements, and postures that resemble someone holding a weapon. They are trained on vast datasets of images and videos.

However, the Maryland Office of the Inspector General for Education later investigated and reported in November 2025 that the Omnilert system actually flagged the student's hand movement, including its position and angle, as a potential threat, not the chip bag itself. It wasn't the Doritos, but how they were held.

Think of it like an automated quality control system at a biscuit factory. It's programmed to spot a broken biscuit by its irregular shape. If a perfectly good biscuit, perhaps slightly tilted on the conveyor belt, casts a shadow that makes it look 'broken' to the machine, it gets flagged. The AI is a pattern recogniser; if a chip bag held in a certain way resembles a gun's profile or a threatening gesture it was trained on, it will flag it. The '90% confidence' is the system's internal probability, not a guarantee of real-world accuracy.

And that brings us to the second question.

Question 2

How Did Human Actions Amplify a False AI Alert into a Police Incident?

This is where the story takes a critical turn. Omnilert's system did generate an alert, and a human review team at Omnilert initially verified it. However, within two minutes, three officials reviewed the image, determined it was a mistake, and cancelled the alert, as per Google Search research notes.

But here's the crucial breakdown: Kenwood Principal Catherine "Kate" Smith opened the original alert over 10 minutes later. Critically, she did not acknowledge the cancellation. Instead, she sent a screenshot of the original, now-cancelled, alert to a school resource officer, who then contacted the Baltimore County Police Department.

At approximately 7:21 p.m., six police cruisers had surrounded the teenagers. Officers unholstered their weapons, ordered the students to their knees, handcuffed them, and conducted a search, according to the lawsuit. Body-worn camera footage obtained by CBS News Baltimore corroborates officers approaching and searching the students before realising the object was a bag of chips.

For context, this sequence shows a complete failure in communication and protocol. An officer allegedly told the students as he left: "Hey, thanks for not running away, that would have been a real problem." Another officer reportedly said, "fucking Omnilerts" and "stupid AI cameras," while a colleague joked, "Be careful on school property, especially holding stuff like that … be careful with your chips bags."

Imagine a cab fleet using an automated system to flag drivers for speeding. The system issues an alert, but the fleet manager's team quickly reviews the GPS data, finds it was a false reading due to a GPS glitch, and cancels the alert. However, a supervisor (the principal) ignores the 'cancelled' status, sees the original speeding alert from 10 minutes ago, and calls the traffic police without checking the updated status. The system worked, the initial human review worked, but the final human in the loop failed to follow the process, escalating a non-issue into a major problem.

The lawsuit also alleges the principal made light of the incident in texts with fellow administrators, including emojis of a smiley face crying with laughter, and did not apologise to the students. This highlights a severe lack of accountability and empathy from the adults involved.

And that brings us to the third question.

Question 3

What Does This Incident Mean for Businesses Deploying AI Security Systems?

The four students are seeking compensatory and punitive damages, along with changes to school and police procedures governing AI-generated threats. The lawsuit alleges violations of the Fourth and 14th Amendments (protections against unreasonable searches and seizures, and due process), Maryland constitutional protections, false imprisonment, and negligence.

Damon Hewitt, president of the Lawyers' Committee for Civil Rights Under Law, representing the students, stated, "These students were failed not only by a technology that is prone to false alerts, but also by the very adults meant to keep them safe when we send them to school." Baltimore County Councilman Julian Jones questioned, "How did it come to be that we had police officers with guns drawn approaching a kid because of a bag of Doritos?"

For context, this incident underscores the profound legal, ethical, and reputational risks involved when AI systems are deployed without robust human oversight and clear operational protocols. It's not just about the AI's accuracy; it's about the entire ecosystem surrounding its use.

Deploying AI is like entrusting a critical task to a brilliant but socially awkward new employee. They can process information at lightning speed but might misinterpret social cues or make errors a human would instantly correct. You wouldn't let them handle sensitive customer interactions without clear guidelines, supervisory checks, and a strong feedback loop. Similarly, AI systems, especially in high-stakes environments like security, need comprehensive human-in-the-loop processes. This includes immediate, qualified human verification of all alerts, clear protocols for escalation and de-escalation, and mandatory training for all personnel on how to interpret and act on AI outputs, including acknowledging cancellations.

Businesses must recognise that AI is a tool, not a replacement for human judgment and accountability. The "technological and human errors" cited in the lawsuit highlight that the human element, from the principal's delayed action to the officers' response, was as critical to the escalation as the AI's initial misidentification. For any business considering AI deployment, especially in areas with potential for human impact, this is a stark lesson in process design, training, and ethical responsibility.

The Bigger Picture

The Doritos incident is more than a cautionary tale about mistaken identity; it's a stark reminder that technology, no matter how advanced, is only as good as the systems and people managing it. It reveals the complex interplay between machine intelligence and human judgment, where a glitch in one can be catastrophically amplified by failures in the other.

For businesses rushing to integrate AI, the lesson is clear: robust technology needs equally robust human processes, accountability, and empathy. Otherwise, a simple bag of chips can turn into a legal and ethical nightmare, proving that the "human touch" remains indispensable, even in an AI-driven world.

Quick Answers

What caused the AI gun alert in the Doritos incident?

The Omnilert AI system flagged a student's hand movement, including its position and angle, as a potential threat, not the Doritos bag itself, according to an investigation by the Maryland Office of the Inspector General for Education.

Was the AI alert immediately cancelled?

Yes, Omnilert's human review team initially verified the alert but then, within two minutes, three officials reviewed the image, determined it was a mistake, and cancelled the alert.

What role did human error play in the incident?

Human error was critical. The school principal opened the original alert over 10 minutes later, did not acknowledge its cancellation, and escalated it to police, leading to the students' unlawful detention.

Sources

Facts are credited to the outlets above. The analysis, calculations and Signal Check are by the Finnect AI Insights desk.