Could AI Hack Your Smart TV? A Live Warning Number That Keeps the Film Playing and Moves Only the Strange App Aside
Abstract
A smart television is a small computer in the living room, with a network connection, often a microphone and sometimes a camera. In October 2026 owners began using AI coding agents to open their own televisions, which shows how open these devices have become and raises the question of what protects a television from someone who uses the same openness against its owner. Today's protections are fixed: software updates, a separate network, and switching off the microphone and camera by hand. This concept note proposes a live layer on top of them. The television, or the home router, watches the behaviour of each app: the pattern of its network connections and its use of the microphone and camera. A lag-1 autocorrelation warning number is calculated at run time, and the computer decides a graded shunting action from it, with no person to wait for. Only the one app that behaves strangely is moved to a side track, with its network cut or its microphone and camera shut, while the film keeps playing. This is a concept to be tested on real device traffic, not a result. Patent pending.
