A Baltimore County high school student was detained at gunpoint after an artificial intelligence gun detection system mistakenly flagged a crumpled Doritos bag in his pocket as a firearm—an error now fueling widespread concern about the reliability and human cost of AI surveillance in schools.
On the evening of October 20, 16-year-old Taki Allen stood outside Kenwood High School after football practice, chatting with friends and eating Doritos. After finishing the snack, Allen stuffed the shiny, crumpled chip bag into his pocket. Moments later, several armed police officers arrived, ordering him to the ground.
“They told me not to move,” Allen recalled. “I didn’t know what was happening. I just froze.” Officers handcuffed and searched him, finding only the chip bag. He was released after several minutes but says the incident left him shaken and afraid to return to school.
“I thought I was going to die over a bag of chips,” Allen said, describing the lasting fear and distrust the experience left behind. “Nobody from school has even talked to me since.”
The chain of events began when Omnilert, an AI-powered gun detection system installed across 7,000 cameras in Baltimore County Public Schools, identified what it believed was a firearm in surveillance footage. The system automatically generated a “gun alert,” prompting school security staff to contact law enforcement before any manual review occurred.
Omnilert, implemented in 2023 as part of a districtwide safety initiative, is designed to rapidly analyze video feeds and notify authorities when it detects potential weapons. Yet, in this case, the system confused the reflective chip bag for a handgun.
AI-driven detection tools like Omnilert are marketed as high-speed solutions to prevent shootings, promising near-instant recognition of firearms. But critics warn that such systems often fail in real-world conditions, especially in schools.
False positives—triggered by harmless objects like phones, binders, umbrellas, or, in this case, a snack bag—are reportedly common. Some systems used in U.S. schools have reported false alarm rates as high as 60%, leading to costly and dangerous police responses.
“These algorithms are only as good as the data they’re trained on,” said Dr. Marcia Reyes, an AI ethics researcher at Johns Hopkins University. “If they aren’t trained to recognize everyday objects in diverse school environments, mistakes like this are inevitable—and they can have terrifying consequences.”
Omnilert and other companies insist their tools save lives by alerting authorities in seconds. However, independent audits have found limited evidence that such systems prevent violence or improve response times. In some cases, AI surveillance has missed real weapons entirely, such as during the 2025 Nashville school shooting, when cameras failed to detect a gun before the tragedy unfolded.
Following the Baltimore incident, parents and students expressed outrage over the AI malfunction and the school’s handling of the aftermath. While the district issued a generic counseling message, Allen said no one contacted him personally.
Community advocates call for an independent review of Omnilert’s deployment, emphasizing the emotional toll such errors inflict. “These systems are supposed to protect kids,” said Renee Walker, a local parent activist. “Instead, they’re putting them in danger and eroding trust between students, schools, and police.”
Across the country, districts have spent millions of dollars on AI-based detection systems amid growing fears of school shootings. But critics say these investments amount to “security theater”—costly measures that project safety without delivering it.
The Federal Trade Commission has reportedly begun investigating companies that market gun-detection AI over concerns of deceptive claims about accuracy and performance.
School officials nationwide are rethinking their reliance on automated systems. “No single tool can guarantee complete safety,” Baltimore County Public Schools said in a brief statement. “We are reviewing this incident to determine how protocols can be improved.”
For students like Allen, that review cannot come soon enough. “I just want to feel safe again,” he said. “If they’re going to use technology like this, they need to make sure it doesn’t ruin someone’s life over a mistake.”
The incident serves as a sobering reminder of the double-edged role of AI in school security—where the promise of safety collides with the peril of human harm.
