Detection equipment to prevent AI exam cheating will not be ready for South Korea's CSAT
Detectors that could stop AI exam cheating won’t be ready for CSAT
Equipment capable of detecting AI exam cheating will not be ready for this Thursday's South Korean College Scholastic Ability Test (CSAT), and experts consider the most realistic approach this year to be proctors visually inspecting candidates' glasses and attaching marker stickers. Previously, a 34-year-old surnamed Han, who ranked first in this year's tax accountant exam, admitted in a YouTube video that he used a miniature camera to photograph questions and send them to AI, then checked the answers on a smartphone. Experts say the radio signal sent when detection equipment transmits questions to AI is the most reliable; the Korea Communications Agency (KCA) called for bids on November 2 to develop a handheld detector that picks up smart glasses' Bluetooth and Wi-Fi signals; blanket signal blocking is difficult to adopt because it lacks a legal basis under the Radio Waves Act and is ineffective against on-device AI that does not need connectivity. A Duke University team proposed in March identifying smart glasses by photographing light leakage from the lenses' waveguides, with an 80% recognition rate; thermal imaging approaches are questioned because of interference from body heat and cost. KAIST professor Yoon Sang-ho calls for adding security by design at the manufacturing stage, such as shutting down when the camera LED is covered and preventing modified devices from operating.