A security researcher says the Samsung Smart TV AI tracking built into his own television goes well beyond basic viewing data, allegedly using connected devices’ microphones to capture audio, taking periodic screenshots of what’s on screen, and sending the results back to Samsung for ad targeting. The claim comes from a bug bounty hunter who reverse-engineered the firmware on his own set, and it has reignited long-running concerns about exactly how much smart TVs listen, watch, and report on the households they sit in.
Key takeaways
- ✓A security researcher reverse-engineered his own Samsung Smart TV and found an onboard AI model
- ✓The code allegedly taps connected devices' microphones to capture audio for ad targeting
- ✓The TV reportedly takes periodic screenshots of on-screen content as part of the system
- ✓OpenAI banned the researcher's account for 'cyber abuse' while he investigated the code
- ✓Samsung has not publicly confirmed or denied the findings
The researcher published his findings after digging through the code running on his Samsung television, describing an onboard AI model that behaves less like a content recommendation tool and more like a quiet surveillance layer. According to his account, the system can draw on microphones belonging to other devices connected to the same network, not just the TV’s own hardware, to pick up ambient audio. That audio, along with screenshots taken while the TV is in use, reportedly feeds into a pipeline built to sharpen the advertising shown to the household.
Table of Contents
What the researcher found inside the TV’s code
Bug bounty hunters spend their time hunting for exactly this kind of behaviour: code that does more than a company’s marketing or privacy policy suggests. In this case, the researcher says he found functions tied to an AI model capable of processing audio input and still-image capture, both of which point toward continuous or near-continuous monitoring rather than a one-off diagnostic feature.
The screenshots are particularly notable because they imply the system can see what’s actually being watched or played on the television, not just metadata about app usage. Combined with audio capture, the picture that emerges is of a device building a detailed profile of what happens in the room it sits in, ostensibly so Samsung and its advertising partners can target viewers more precisely.
None of this has been independently verified by Samsung, and the company has not issued a public statement confirming or denying the specific claims. That matters, because reverse-engineering firmware is technically demanding and open to misinterpretation, even when the researcher doing it is experienced. Still, the level of detail in the published findings, down to specific code functions tied to audio and image capture, has been enough to worry security-conscious owners of Samsung televisions.
How the alleged ad-targeting system is said to work
The researcher’s description suggests a layered system rather than a single feature. Audio captured through connected microphones, screenshots of on-screen content, and other usage data are allegedly combined to build a advertising profile unique to each household’s viewing habits.
| Alleged data collected | Stated or implied purpose |
|---|---|
| Audio via connected device microphones | Ad personalisation based on ambient sound |
| Periodic screenshots of on-screen content | Matching ads to what’s being watched |
| Other usage and device data | Building targeted advertising profiles |
This isn’t the first time smart TV makers have faced scrutiny over automatic content recognition and data collection built into their platforms. What sets this claim apart is the alleged use of other connected devices’ microphones rather than just the TV’s own hardware, which, if accurate, widens the scope of what a single television could theoretically capture across a whole household’s network.
Banned by OpenAI during the investigation
While working through the reverse-engineering process, the researcher says his account with OpenAI was banned for violating its policy against cyber abuse. He had reportedly been using AI tools to help analyse and make sense of the TV’s code, a common technique among security researchers trying to parse large, obfuscated firmware. OpenAI’s usage policies explicitly restrict using its models to assist with unauthorised access to systems or data, and automated enforcement systems can flag legitimate security research in the same way they flag malicious activity.
The irony hasn’t been lost on observers: a researcher investigating one company’s alleged use of AI to quietly collect household data found himself penalised by another AI provider for using similar tools to do the investigating. It’s a reminder that the rules governing what counts as acceptable use of AI assistance are still being worked out, often clumsily, even as AI models themselves get embedded deeper into everyday hardware like televisions.
What it means for smart TV owners
For anyone with a Samsung Smart TV, or any internet-connected television for that matter, the findings are a useful prompt to check privacy and data settings rather than a confirmed verdict on what the company is doing. Most smart TV platforms include options to limit automatic content recognition, disable voice features, and restrict ad personalisation, even if those settings are often buried several menus deep.
The broader implication is bigger than one brand. Televisions have quietly become some of the most data-hungry devices in the home, sitting on the network alongside phones, laptops, and streaming boxes, often with far less scrutiny than those other devices receive. The idea of an AI model quietly processing audio in the background isn’t unique to this story either; Apple’s recent acquisition of AI audio editing startup Samplab shows how aggressively the industry is building AI into audio pipelines, albeit for very different purposes than ad targeting.
Until Samsung responds directly to the specific claims, owners concerned about what their television might be collecting have a few practical options: reviewing privacy settings, disconnecting unnecessary smart features, and being deliberate about which devices share a network with a smart TV. Readers already taking similar precautions elsewhere in their home theatre setup might find a guide on using a VPN with Plex to keep streaming habits private covers related ground worth considering.
Frequently asked questions
Has Samsung confirmed or denied the claims?
No official statement from Samsung directly addressing these specific findings has been published. The claims currently rest on the researcher’s own reverse-engineering work and have not been independently verified by the company or other security outlets.
Can Samsung Smart TV owners turn off this kind of data collection?
Samsung televisions typically include settings to limit automatic content recognition, disable voice assistant features, and restrict personalised advertising, though these options are often located several layers deep in the settings menu rather than presented upfront.
Why was the researcher banned from OpenAI?
He says his account was suspended for violating OpenAI’s policy against cyber abuse while he was using AI tools to help analyse the television’s code, a common research technique that appears to have been flagged by automated enforcement systems.
Is this the first time smart TVs have faced privacy concerns?
No. Smart TV platforms across multiple brands have previously faced criticism over automatic content recognition and data sharing with advertisers, though the alleged use of other connected devices’ microphones in this case goes further than most prior reports.
