Security researchers have uncovered a disturbing finding: LG smart TVs continue collecting user data even when owners believe their devices are offline or powered down. The discovery, first reported by The Verge, reveals that background processes remain active in standby mode, logging everything from viewing habits to voice commands—all feeding into AI training datasets without explicit user consent.
This isn't just a minor privacy hiccup. For content creators who use smart TVs to review their work, test audience experiences, or simply unwind, the implications are immediate: your viewing patterns, the content you create and consume, and even casual voice interactions are potentially being harvested to train AI models you'll never see or control.
The Discovery: Data Collection Never Stops
Independent security researchers analyzing network traffic from LG smart TVs discovered persistent data transmission patterns that continued even after users had disabled all networking features through the TV's settings menu. The devices maintain what LG internally calls "low-power monitoring mode"—a state that keeps specific components active while the main display remains off.
The research team identified data packets being stored locally on the TV's internal memory during offline periods, then automatically uploaded in batches whenever the device detected an available internet connection. This could happen through Wi-Fi reconnection, Ethernet cables plugged in for firmware updates, or even neighboring smart home devices that create ad-hoc networks.
Even physically disconnecting your LG TV from the internet doesn't stop data collection—it just delays the upload.
Models affected include LG's OLED C4, G4, and Z4 series, along with their QNED and NanoCell ranges manufactured from 2024 onwards. LG estimates approximately 12 million units worldwide fall into this category, though the company hasn't confirmed whether older models contain similar functionality.
What Data Is Being Collected
The scope of data collection is broader than most users realize. According to the research findings, LG TVs log:
Content metadata: Titles, timestamps, and duration of everything watched—whether through streaming apps, HDMI inputs, or broadcast TV. This includes your YouTube uploads if you're reviewing them on the big screen, client presentations loaded via USB, or any video content played through connected devices.
Voice interactions: Full transcripts of voice commands issued through LG's ThinQ AI remote, including failed recognition attempts and background ambient audio during active listening periods. The system captures roughly 3-5 seconds of audio before and after each detected wake word.
Usage patterns: Navigation behavior within smart TV interfaces, app launch frequencies, settings adjustments, and even cursor movement during text entry. This creates detailed behavioral profiles that can identify individuals within households based purely on interaction style.
Network environment data: Information about nearby Wi-Fi networks (even those you haven't connected to), Bluetooth devices in range, and connected smart home products. This metadata helps AI models understand contextual usage environments.
The AI Training Connection
Where does all this data go? According to LG's updated privacy policy (buried in a September 2026 revision), the company uses aggregated viewing data to "enhance AI-powered content recommendation systems and improve voice recognition accuracy across our product ecosystem."
Translation: your watching habits are training LG's recommendation algorithms and natural language models. While LG insists data is anonymized before use in model training, security experts point out that the sheer volume of behavioral data makes true anonymization nearly impossible. Viewing patterns combined with voice characteristics create unique fingerprints that can often be traced back to individuals.
- Behavioral Fingerprinting
- The practice of creating unique user identifiers from patterns of interaction rather than explicit personal information. Even without names or email addresses, combinations of viewing times, content preferences, and navigation habits can identify specific individuals with 87% accuracy across datasets.
This matters especially for creators because the content you produce and consume becomes training data for AI systems that may eventually compete with your work. If you're a video editor reviewing client projects on your LG TV, that content metadata enters the training pipeline. If you're a voice actor practicing scripts through the TV's casting features, your vocal patterns help train voice synthesis models.
The commercial implications extend further: LG has partnerships with multiple AI companies, including undisclosed deals with content recommendation platforms and advertising technology firms. While LG won't confirm specifics, industry analysts believe collected data is being licensed to third parties for additional AI training purposes.
What This Means for Creators and Users
For the average viewer, this is invasive. For content creators, it's potentially business-impacting. Consider these scenarios:
You're a YouTube creator reviewing your latest upload before publishing. Your LG TV logs the content, timestamps, and how many times you rewatch specific sections. This metadata—showing what creators focus on during review—becomes training data for AI systems learning to predict successful content patterns.
You're a filmmaker evaluating color grading on an OLED display. Your adjustments to picture settings, the specific scenes you linger on, and even your voice commands requesting brightness changes all feed into datasets about professional color workflows.
Content Strategy
Your private content reviews become public training data for recommendation algorithms
Voice Work
Practice sessions and vocal warm-ups train competing voice synthesis models
Client Work
Commercial projects viewed for QA purposes enter AI training pipelines without client consent
Privacy Control
Existing privacy settings don't actually prevent data collection as advertised
The practical workaround options are limited and inconvenient. You can physically disconnect the TV from all networks—but this disables legitimate smart features like streaming apps and firmware updates. You can disable voice recognition—but background audio sampling continues for "quality assurance purposes." You can opt out of personalized advertising—but LG's privacy policy explicitly states this "does not affect data collection for product improvement and AI development."
Several content creator communities are now recommending using LG TVs exclusively as "dumb" displays, connected only via HDMI to external media players with their own privacy controls. This defeats the purpose of buying a smart TV, but it's currently the only way to ensure true data isolation.
LG's Response and Next Steps
LG Electronics issued a statement acknowledging the data collection practices but defending them as "industry standard for modern smart home products." The company emphasizes that collected data is "anonymized, encrypted, and used solely to improve customer experience through better AI-powered features."
However, LG committed to three specific changes in response to the security research:
Updated privacy controls: A firmware update rolling out this month will add a master "Data Collection Off" toggle that genuinely disables all background logging when activated. LG claims this will result in "degraded smart features" but won't affect basic TV functionality.
Transparency dashboard: LG will launch a web portal where users can view exactly what data their specific TV has collected and request deletion of stored information. The portal goes live in October 2026, though accessing it requires creating an LG account—which itself involves agreeing to separate data policies.
Before
Privacy settings claimed to disable data collection but background processes continued logging activity
After
New master toggle genuinely stops data collection but disables smart features entirely
Before
No visibility into what data was collected or how it was used
After
Transparency dashboard shows collected data and allows deletion requests
Before
Offline mode still allowed local data storage for later upload
After
True offline mode deletes locally stored data if not uploaded within 72 hours
Opt-in AI training: Starting with 2027 TV models, LG will make AI training data contribution an opt-in choice during initial setup rather than an opt-out buried in settings. Existing TV owners will receive a one-time prompt with the October firmware update asking them to explicitly consent to continued data collection.
Privacy advocates are skeptical these changes go far enough. The Electronic Frontier Foundation noted that "requiring users to choose between privacy and basic smart TV functionality creates a false dichotomy—manufacturers should be capable of delivering features without pervasive data harvesting."
For creators, the immediate takeaway is simple: audit your smart TV settings now. Check what you've actually consented to versus what you thought you agreed to. And consider whether the convenience of smart features is worth becoming unpaid training data for AI systems that may eventually compete with your creative work.
The broader pattern is troubling. As AI companies scramble for training data, consumer devices are increasingly becoming covert data collection endpoints. Your TV is just the beginning—smart speakers, home security cameras, even fitness trackers are all potential training data sources. The LG discovery is less an isolated incident and more a preview of how consumer electronics companies are monetizing our behavioral patterns in the AI era.