Get Full Government Meeting Transcripts, Videos, & Alerts Forever!
Get email alerts on the School Security Ai topic
No spam. Unsubscribe anytime.
Board hears state-funded AI camera pilot demo; questions focus on privacy, detection limits and long-term costs
Summary
Vendors Evident and Raptor presented a proposal to pilot an on-premise computer-vision AI system for campus detection and automated alerts at selected New Hanover schools. Board members asked about facial recognition, license-plate reads, false positives, cloud vs. edge processing, state funding duration and ongoing licensing costs.
Vendors Evident and Raptor demonstrated an on-premise computer-vision artificial intelligence system in a board meeting presentation, one of the items included for information and planned as a state-supported pilot. Jared Worthington, the district’s executive director of school support, introduced the restart- and pilot-related presentation; Evident and Raptor representatives described threat, smoke, slip-and-fall, loitering and weapon-detection use cases and how alerts would route to the Raptor emergency-management platform.
Presenters said the proposed design uses “edge” processing: a local server at each school runs AI models against live camera streams and sends only event alerts (not continuous video or personally identifying records) to the notification system. Vendors said the system can be licensed for a pilot period (they proposed a two-year licensing/support window for the district) and that ongoing licensing/support after the pilot would be roughly 20 percent of the software cost per year; hardware would be purchased up front.
Board members asked detailed questions about privacy and operational limits: Dr. Tim Merrick raised concerns about facial recognition and profiling; vendors said the system is not configured to perform broad facial identification unless the district chooses to load a vetted list (for example, registered offenders) and enable that capability. Vendors said visual detection does not identify concealed weapons until a weapon is in view; they described milliseconds-long detection-to-notification times once a weapon is visually detected. The vendors said license-plate recognition can be configured to look for specific plates, rather than recording all plates.
Costs and funding were discussed: presenters said the state allocated funds for the first two years of pilots; the vendor estimated district out-year licensing/support at roughly 20 percent of initial software costs (vendors gave an order-of-magnitude estimate in the meeting). Participants pressed vendors on false positive/false negative rates; vendors replied that accuracy varies by camera model, angle and lighting and that administrators can tune rule sets and retrain models using false alerts.
The board extended the presentation time during Q&A and directed staff to follow up. No procurement or contract was approved at the meeting; staff will bring details to future agenda items and consent packets for board consideration.
