A reverse engineer’s investigation into a 2025 Samsung smart TV has raised uncomfortable questions about how much data modern televisions may process in the name of personalized advertising. The researcher, known online as log4j, says they found Tizen code associated with detailed telemetry, audio categorization, screenshots, viewing behavior, and cross-device advertising identifiers—claims detailed in their investigation of Samsung’s TV software.
The inquiry began after log4j noticed what they considered unusually specific advertisements appearing on the television. After gaining a root shell and examining the encrypted firmware through hardware-level techniques, they said they discovered background systems that could analyze network activity, identify viewing patterns, and process images from the screen. Other coverage of the findings describes the alleged system as including audio analysis, screenshot capture, and an advertising identifier that could help recognize a user across connected devices. The findings remain the researcher’s interpretation of reverse-engineered code, rather than an independently reproduced security audit.
One especially striking claim involves machine-learning software embedded in the Tizen environment. Log4j says they found a region-specific convolutional neural network that could help classify behavior such as which videos someone watches, which advertisements they skip, and how far they progress through a video. They also suggested that the processing burden might help explain why some smart TVs feel sluggish even when users have not installed demanding applications. That does not establish that every Samsung television performs these operations continuously, but it illustrates how advertising technology can move from remote servers onto consumer hardware.
The microphone question is murkier. According to the researcher, even a television without its own microphone appeared capable of looking for connected hardware—such as a soundbar or Bluetooth accessory—that might contain one. A separate observer, identified as ngxson, said a rooted Tizen television opened a microphone for roughly five seconds before stopping. However, that test was performed on a modified device with multiple settings disabled, so it cannot determine how an unmodified retail television behaves.
Samsung’s privacy documentation confirms that its ecosystem can use device identifiers for Smart TVs, including a Personalized Service ID and a Tizen Identifier for Advertising. The policy also describes usage and log information, personalized advertising, technologies that measure whether content or ads have been viewed, and analytics across connected devices and services. It says voice recordings are collected when users enable voice-command functions, but the policy page does not specifically confirm the alleged screenshot pipeline or the broader microphone behavior described by log4j. Samsung’s privacy policy also says users may have rights to request access to, correction of, or deletion of personal information, subject to legal exceptions.
The episode also became entangled with a separate dispute involving OpenAI. Log4j said they used OpenAI tools for local reverse-engineering work rather than to discover or exploit vulnerabilities, but later received an account-ban message citing cyber abuse. They argued that an automated classifier had misunderstood the activity. For now, the Samsung allegations point to an important distinction for smart-TV owners: a privacy policy may disclose broad categories of collection, while the actual behavior of an individual model depends on its firmware, connected devices, region, account settings, and enabled features. A fuller technical write-up and independent testing will be needed before the most serious claims can be treated as established behavior across Samsung’s television lineup.
