WORLDTECHBounty hunter says Samsung TVs record audio and screenshots for ad targeting
A cybersecurity bounty hunter says reverse engineering a Samsung TV exposed Tizen code that collects user data, including audio and screenshots, to make targeted advertising more precise. The claims come from one researcher's public posts, as reported by Tom's Hardware.
What the researcher found
The reverse engineer and bounty hunter log4j had been receiving unusually specific advertisements on their Samsung TV, which led them to examine the device's software more closely. Working through the Tizen operating system and posting results as they went, log4j described code that draws detailed telemetry from packet logging, sorts audio, takes screenshots and links activity across devices for ad tracking.
According to the account, this material is meant to extract as much information about each user as possible in order to refine advertising. The researcher also flagged a region-specific convolutional neural network, an AI model that monitors viewing habits, among them which videos are watched, which advertisements are skipped and how far playback has progressed.
Microphone behaviour and an independent check
log4j further reported logic for televisions without built-in microphones, saying such a set would look for connected devices with microphones, such as a soundbar or a paired Bluetooth device, in order to record audio.
Hugging Face engineer ngxson said they tested this with a Flipper Zero and observed the TV open its microphone for five seconds before it stopped. A later check brought no further microphone activity. ngxson added that their Tizen TV was already rooted and had several settings disabled, so an unmodified set may behave differently.
OpenAI account ban
log4j said the research involved OpenAI tools and that the company later banned the account for cyber abuse, as an email to the researcher stated. log4j rejected the suggestion that the tools were used to exploit or find vulnerabilities, saying they served only local reverse engineering and that the terms of use had not been broken. In a comment on the ban, log4j suggested a primitive classifier may have been responsible.
What is confirmed so far
The findings rest on the researcher's own reverse engineering and public posts. The reporting includes no independent audit of the Tizen code and no official response from Samsung, and the single microphone test did not show the behaviour repeating.
Open questions
- Will Samsung respond to the findings on audio, screenshots and ad tracking?
- Does the microphone behaviour also occur on unmodified Samsung TVs?
- What led OpenAI to ban the researcher's account?