Samsung running a TensorFlow lite model on 960x540 image to track you

Samsung is running a TensorFlow Lite model called "alex.tflite" (based on AlexNet) on its devices to identify games users are playing by analyzing 960×540 images with INT8 quantization every few seconds. The model performs poorly with a 99% failure rate, wastes CPU resources, and appears to process data before requesting user consent. This discovery represents a significant privacy and performance concern on Samsung's Tizen platform.