A bag of chips got a teen handcuffed. Here's what AI security systems keep getting wrong in schools.
A student walked into a Baltimore County school with a snack. Minutes later, they were in handcuffs on the floor while police sorted out that the "weapon" was a bag of chips.
NBC News reported the incident on October 28, 2025, highlighting how an AI-powered security system flagged the teen's bag as containing a firearm. Officers responded to the alert and restrained the student before the error became obvious. No charges. No weapon. Just a traumatized kid and a school district with some explaining to do.
Why AI false positives in schools are a serious problem
AI weapons detection systems have been marketed aggressively to K-12 districts over the last three years, often as a faster, less invasive alternative to metal detectors. The pitch is real: these systems can scan hundreds of students per minute using computer vision without slowing down the entry flow. But the false positive rates are rarely front-and-center in the sales conversations.
Research from the Georgetown Center on Privacy and Technology found that AI surveillance tools deployed in schools have disproportionately misidentified students of color at higher rates than their peers. While the Baltimore County incident has not been characterized in those terms publicly, the pattern matters. When a system cries wolf, the consequences are not abstract. A real student gets detained. A real response gets triggered.
Across the US, schools have spent an estimated $3.1 billion on security technology since 2018, according to the K-12 Security Information Exchange. Much of that money has gone into systems that have never been independently audited for accuracy under real-world conditions.
The human layer is not optional
This is exactly the kind of situation XGuard handles in its deployment protocols. AI-assisted detection is a tool, not a decision-maker. XGuard operators are trained to treat any automated alert as a prompt to investigate, not an instruction to act. The workflow is alert, assess, confirm, respond. Skipping the middle two steps is how a bag of chips becomes a use-of-force incident.
In the Baltimore case, the failure was not necessarily the AI itself. It was the response protocol. If the system flagged a potential threat and officers moved to physical restraint without a human verification step, the system design failed the student just as much as the algorithm did. A trained guard reviewing camera footage or conducting a quick visual check before escalation would likely have caught the error in under thirty seconds.
The Northwestern University dorm attack reported the same week offers a useful contrast. In that case, three students physically intervened to stop a violent attacker and protect a classmate. No AI, no detection system. Just people paying attention and acting with judgment. The two stories together make a point that security technology vendors rarely want to make: situational awareness and human decision-making remain irreplaceable.
What schools should demand from their security vendors
If your district is using or considering AI detection technology, there are a few things worth pushing for before the next school board meeting.
First, ask for the false positive rate under conditions that match your environment. Cafeteria lighting, crowded hallways, students carrying bulky backpacks and yes, bags of chips. Lab conditions and real-world conditions are different.
Second, make sure your response protocol requires human confirmation before any physical intervention. An alert should start a process, not end one.
Third, review who gets notified when a false positive occurs and what the remediation looks like. If the answer is "nothing formally," that is a liability waiting to happen.
Pro tip: Any AI security alert at a school entrance should trigger a two-step verification: a human operator reviews the flagged image or live feed before any physical response is authorized. Build that pause into your post-orders. It takes fifteen seconds and prevents incidents that take fifteen months to litigate.
The trust cost nobody budgets for
False positives do not just embarrass a vendor. They erode student trust in the institutions supposed to protect them. A teenager who gets handcuffed in front of classmates for carrying a snack does not forget that. Neither does anyone who watched it happen.
School security is not just about stopping threats. It is about creating an environment where students feel safe enough to learn. Those two goals are not in tension, but they do require security systems, human and technological, that are accountable, accurate, and supervised.
The technology will keep improving. The human oversight layer needs to be there in the meantime.
Need protection where you are? XGuard connects you with licensed, vetted security operators in minutes — for events, residences, retail, executive protection, and fire watch. Available globally.
Source: us-nbc-news — 2025-10-28
Published by XGuard, the on-demand security marketplace.