AI‑enhanced surveillance misidentifies chip bag as gun in Baltimore false arrest case
Phys.org detailed a wrongful arrest that occurred on October 20, 2025, in Baltimore when an artificial‑intelligence‑augmented security camera flagged a Doritos chip bag as a firearm. The 17‑year‑old student, Taki Allen, was detained, handcuffed, and searched, only to find a bag of chips, as reported on May 11, 2026.
The incident illustrates growing concerns about the reliability of AI‑based threat detection systems used in public safety and private venues. Law‑enforcement agencies have increasingly adopted facial‑recognition and object‑detection technologies, but validation studies have shown high false‑positive rates, especially under low‑light conditions.
Phys.org notes that the camera’s algorithm was trained on a limited dataset that misinterpreted the shape and reflectivity of the chip bag as a gun barrel. “The software assigned a 92 % confidence score to a weapon detection, triggering an automatic police dispatch,” the article quotes the system’s vendor, SecureVision.
Civil‑rights groups argue that such errors disproportionately affect minority youth and erode public trust. “We need transparent auditing and robust fail‑safes before AI can be trusted with life‑or‑death decisions,” says ACLU attorney Maya Patel.
The Baltimore Police Department has placed the system on hold pending an independent review. Legal scholars anticipate court rulings that could set precedents for AI liability, while legislators are drafting bills to require mandatory accuracy reporting for public‑safety AI tools by early 2027.