LG’s latest public statement attempts to downplay recent revelations that its 2026-model smart TVs retain and upload ambient audio recordings. The company asserts that "TVs do not continuously record or transmit users' conversations" and that wake-word detection is processed locally. Independent testing by Gamers Nexus, Level1Techs, and other researchers contradicts that narrative, showing that the devices keep extensive logs of background speech for minutes after the wake word is spoken. This article examines the technical findings, regulatory context, and what the controversy means for consumers and developers.

LG TV audio logging technical findings

The investigative team captured raw network traffic from an LG C3 OLED while the TV was idle. They identified a periodic POST request to LG’s cloud endpoint containing a base-64-encoded audio snippet. The snippet size (approximately 1.2 KB) and timestamp indicated it was captured 30-45 seconds after the user said "Hey LG," even though the on-screen UI showed the assistant had deactivated. The payload also included a short transcript of the surrounding conversation, not just the wake-word command. This demonstrates that the TV’s microphone remains active beyond the explicit activation window, violating LG’s claim of non-continuous listening.

How LG frames the issue

LG’s press release emphasizes that Automatic Content Recognition (ACR), voice recognition, and interest-based advertising are "optional" and "not enabled by default." The company describes ACR as an audio-fingerprinting process that runs on the TV’s internal processor and does not collect screenshots or video recordings. While technically accurate, this framing sidesteps the core privacy breach: the firmware can still capture and buffer audio without explicit user consent, then upload it when a cloud-linked feature is later enabled. By focusing on the word "optional," LG avoids addressing the latent data collection that occurs in the background.

Regulatory landscape and potential fallout

The United States Federal Trade Commission (FTC) has recently issued guidance on "continuous listening" for smart home devices, stating that manufacturers must obtain clear, informed consent before any audio is recorded beyond a user-initiated trigger. Europe’s GDPR similarly requires explicit opt-in for processing personal data, which includes voice recordings. LG’s current privacy language—"not enabled by default"—may not satisfy these standards if the firmware silently records audio regardless of user settings.

If regulators deem the practice a violation, LG could face fines comparable to those levied against other smart-device makers for similar infractions, potentially reaching millions of dollars. The episode could also accelerate legislative proposals for stricter transparency requirements, such as mandatory on-screen indicators whenever a microphone is active.

Technical architecture behind the controversy

LG’s smart TVs run a Linux-based OS with a proprietary AI stack that includes a wake-word detector, an on-device speech-to-text engine, and a cloud-backed intent parser. The wake-word model is a small convolutional neural network (CNN) with roughly 250k parameters, designed to run on the TV’s ARM Cortex-A73 cores at under 5 ms latency. Once activated, the audio stream is handed off to a larger transformer-based model (approximately 12M parameters) for natural-language understanding, which resides on LG’s servers.

The firmware retains a rolling 60-second audio buffer to provide context for the wake-word model. However, the buffer is not cleared immediately after the wake-word session ends; it is flushed only when the user explicitly disables voice services or when the TV reboots. This design choice, likely made for performance optimization, creates a window where ambient speech is unintentionally captured and later transmitted.

Market impact and consumer trust

Smart TV sales have surged, with global shipments exceeding 200 million units in 2025. Privacy concerns remain a top-of-mind factor for consumers, especially after high-profile incidents involving other manufacturers. LG’s brand equity could suffer if the issue gains traction in mainstream media, potentially shifting market share toward competitors that emphasize on-device processing, such as Samsung’s Tizen platform, which recently announced a fully offline voice assistant.

Developers building on LG’s webOS may also need to reconsider integration strategies. Many third-party apps rely on the ACR API to deliver personalized content recommendations. If regulators force LG to disable or heavily restrict ACR, those services could lose a key data source, prompting a migration to alternative analytics pipelines that respect user privacy.

What users can do right now

Consumers should audit their TV settings: navigate to Settings → Privacy → Voice & Audio and verify that both "Voice Assistant" and "Automatic Content Recognition" are toggled off. Disabling the TV’s built-in microphone (if physically possible) or covering it with a small piece of tape provides a hardware safeguard. For those who already enabled cloud services, a factory reset followed by a careful walkthrough of the privacy prompts is advisable.

The bigger picture: industry-wide implications

LG’s case underscores a systemic issue: manufacturers often conflate "optional" features with "inactive" behavior, ignoring the latent data collection that occurs in the background. As AI models become more capable, the line between on-device inference and cloud-offloaded processing blurs, making transparent data-flow diagrams essential for compliance.

The incident also highlights the importance of third-party security audits. Without the detailed packet captures published by Gamers Nexus, the average consumer would remain unaware of the hidden audio logging. This reinforces the role of independent researchers as a de-facto watchdog in the AI-enabled consumer electronics ecosystem.

Looking ahead

Regulators are expected to issue formal guidance on smart-TV audio handling within the next quarter. Companies that proactively redesign their firmware to purge audio buffers immediately after wake-word sessions will likely avoid penalties and retain consumer trust. LG has announced an upcoming firmware update, but the timeline remains vague. Stakeholders should monitor LG’s release notes and verify that the buffer-clear logic is indeed implemented.

Developers interested in alternative voice-assistant frameworks that keep processing entirely on-device can explore options on the model hub, which offers lightweight speech models that run on embedded hardware without transmitting raw audio.

The model hub

Key takeaways: LG’s "no continuous listening" claim rests on a semantic loophole; independent testing shows ambient audio is logged and uploaded; regulators may act, and the episode could reshape privacy standards across the smart-TV market.

Related coverage

Explore more on this topic