Hacker reverses Samsung TV, finds on-board AI listening OpenAI account suspended

Your Samsung TV may be sending conversations to OpenAI, even with the microphone off.

A security researcher has discovered that certain Samsung TVs are equipped with an onboard AI that continuously listens and sends data to OpenAI. The researcher, who goes by the name taiids, found that the AI, which appears to be a local instance of OpenAI's Whisper model, was activated even when the TV's microphone was supposedly off. This AI processed conversations in the room and sent them to OpenAI for transcription. taiids found that their personal OpenAI account was suspended after this activity was detected, suggesting OpenAI flagged the data transmission as suspicious or against terms of service. The discovery raises significant privacy concerns regarding the continuous data collection by smart home devices.

The presence of a local AI instance on Samsung TVs, capable of real-time audio processing and transmission to OpenAI, indicates a significant shift in how smart home devices handle user data. This goes beyond simple voice command activation, suggesting a more pervasive data collection mechanism is in place. The suspension of the researcher's OpenAI account highlights a potential security or privacy blind spot, implying that such continuous data streams might be flagged by AI providers.

This development signals a concerning trend of 'always-on' data capture by consumer electronics, potentially creating new vectors for privacy breaches. The integration of powerful AI models like Whisper directly into hardware, without explicit user opt-in for continuous background operation, raises questions about consent and data security. Consumers may need to be more vigilant about the data collected by their smart devices and understand the implications of their usage patterns on their privacy and account security.