How North Korea Built a Homegrown AI Facial‑Recognition Smartphone

Recent reports from the North Korean media outlet “Echo of Unity” introduced the self‑developed “Blue Sky” smartphone, which boasts fast processing, face‑recognition and fingerprint unlocking. The face‑recognition module was created by the Kim Il‑sung University’s AI research centre.

North Korea’s domestic network, Kwangmyong, runs on CentOS and provides only an internal intranet; only a few officials and foreign visitors can access the global Internet. Consequently, the country’s AI development occurs in a closed, LAN‑only environment.

Despite these constraints, researchers have built a real‑time facial‑recognition system for ARM‑based devices. The algorithm relies on convolutional neural networks (CNN), with face detection based on MTCNN and recognition using MobileFaceNets. Implementation uses TensorFlow and the ArcFace loss function. Training employed a database of 30 million faces, and the resulting model runs on the Blue Sky phone and other tablets via an Android‑Studio application.

The technology has also been demonstrated in other products, such as AI‑enabled smart speakers that translate 30 subjects into Korean, and AI‑driven production‑management systems for factories and tram manufacturing.

These developments illustrate North Korea’s shift from cloud‑centric AI research to distributed, edge‑focused solutions, even though the nation remains isolated from the global Internet.