Core Definition
Anonymized CCTV analytics is a technology in which AI extracts only behavioral data — such as location, movement, and dwell — from CCTV footage, without capturing or processing personally identifiable information like faces, names, or license plates. It secures the insights needed for store analysis while remaining compliant with privacy regulations.
The Legal Background for Anonymized CCTV
Under Korea's Personal Information Protection Act, identifying an individual through CCTV, or analyzing and storing that footage, requires the data subject's consent. For a store visited by an unspecified public, obtaining that consent individually isn't realistic.
Anonymized AI solves this problem at its root. Because it extracts only "what action" occurred in the footage, not "which person" performed it, it never collects data that qualifies as personal information. This is why stores can gather operational data without legal risk.
The Technical Principles of Anonymization
The AI model detects people in the footage but does not extract facial feature points. Instead, it analyzes only a person's location and movement using silhouettes, skeletal keypoints, or bounding boxes. Because this data consists solely of coordinate values and timestamps, it cannot identify an individual even when linked back to the original footage.
Advanced anonymization systems use real-time edge processing, on the camera or a local server, to strip out all identifiable information before the original footage ever reaches an external server. With SAAI, the original CCTV footage stays only on devices inside the store, and only the encrypted, analyzed behavioral data is sent to the cloud.
How SAAI Uses It
All SAAI products (saai care, saai insight, saai agent) run on an anonymization engine. The moment footage arrives from the camera, it is anonymized, and only behavioral data is stored on SAAI's servers. The system is designed to comply with Personal Information Protection Commission guidelines.
See it in action