The Zoom subsidiary uses deepfake detection based on identifying characteristics like rendering artifacts, along with IP location, answer timing, tab switching and LLM answer matching to detect candidate fraud. The signals are compiled into a single view to enable reviews, investigations and ultimately decisions.