A report about LG smart TVs had reached 792 Hacker News points and 384 comments by MrKeyoor's September 7 snapshot. That rush of attention is useful for one reason: developers saw a network-boundary problem hiding inside a television story. The Hacker News discussion is a measure of concern, though, not proof of the technical claims.
The evidence comes from a 135-minute Gamers Nexus investigation, conducted with Level1Techs and independent security researchers. Tests on retail LG OLED televisions, including the G5, found local-network scanning, voice input retained while the set was disconnected from Ethernet, and an upload after connectivity returned. Notebookcheck's account also says the microphone captured clear audio while the display appeared to be off in standby.
That combination matters more than the familiar fact that connected TVs measure viewing. A device in standby can still observe its room, inventory parts of its network and queue data for later transmission, according to the test. For a home user, that makes the television another active computer on the LAN. For an office, lab or shared accommodation, it raises a harder question: what else can a screen reach after everyone assumes it has been turned off?
The display was dark, but webOS kept working
During bench testing, the team captured microphone audio while the screen looked powered down. The state was standby, according to Notebookcheck, so the finding should not be retold as a television recording while unplugged. That distinction narrows the claim without making it trivial. People commonly leave a television connected to power and the network around the clock.
The offline part of the test is especially useful. After Ethernet was disconnected, the set reportedly continued saving voice input locally. It sent the stored data after the network connection came back. The published screenshots also showed voice prompts logged as plain text, according to the same report. In other words, removing connectivity interrupted transmission during the test, but did not stop local collection.
Packet captures made with Wireshark showed the TVs scanning the local network for other hardware. Gamers Nexus and Notebookcheck describe phones and smartwatches among the detected devices. The report says the televisions gathered internal IP addresses as well as the names and signal strengths of nearby Wi-Fi networks, plus location data.
Those observations cover two different surfaces. A LAN scan can describe devices reachable inside the home or office. Nearby Wi-Fi names and signal levels describe the radio environment outside that LAN as well. The report links the resulting data pool to LG Ad Solutions, LG's targeted-advertising arm, but the public summary does not establish that every field observed in the test was used for the same advertising purpose.
Content recognition explains only part of the traffic
Automatic content recognition, or ACR, is already common in connected televisions. It samples audio or video playing on the screen, turns that sample into a fingerprint and matches it to known content. Notebookcheck says LG's implementation can log material watched across inputs, so using a cable box or an external player does not by itself prevent the television from recognizing what appears on the panel.
Regulators have dealt with this class of tracking before. In 2017, the US Federal Trade Commission and New Jersey reached a $2.2 million settlement with Vizio. The agencies alleged that software on 11 million televisions collected second-by-second viewing data without users' knowledge or consent. The order required prominent disclosure, affirmative consent and deletion of data collected before March 1, 2016.
The LG test goes beyond a conventional ACR dispute because it reports microphone collection in standby and discovery of unrelated network devices. Viewing history describes what crossed the screen. Voice input and a local device map can describe activity around the screen. Combining those categories would make a television a much richer source of household context, although the investigation as summarized does not prove that LG combined every observed data type into individual profiles.
Scale figures need the same restraint. Notebookcheck quotes LG as claiming roughly 216 million smart-TV sales worldwide, while LG Ad Solutions says it can reach 363 million secondary addressable devices in the United States. Those figures explain the 216 million number in the Gamers Nexus title and the advertising business's interest in cross-device data. They do not show that 216 million sets share the tested behavior, or that all 363 million secondary devices were found through the scanning method demonstrated here.
The findings are specific, and the open questions are too
The strongest parts of the report are observable events on retail hardware: packet captures showed network discovery, the microphone produced intelligible audio in standby, and data held during an Ethernet outage left after reconnection. Those results deserve attempts at reproduction. Neither a viral headline nor hundreds of Hacker News comments can substitute for tests across firmware versions, regions and consent choices.
The reporting names the LG G5 and says other retail OLED models were included. It does not turn that sample into a model-by-model finding for LG's entire installed base. Firmware, regional services and the options accepted during setup could change what a particular unit does. A useful response from LG would identify affected models and software versions, then explain which settings control each observed data flow.
The power state also has to stay precise. The screen appeared off, but the television remained powered in standby. The test therefore says something important about ordinary daily use, not about a physically disconnected set. Calling standby fully off would make the account easier to dismiss and would hide the practical issue: a dark display gave no visible sign that collection was continuing.
Gamers Nexus also says the researchers found remote-code-execution flaws in webOS and are taking them through responsible disclosure. Notebookcheck does not publish exploit steps, affected versions or patch information. The vulnerability work should remain separate from the telemetry findings until the researchers or LG release enough detail to assess exposure.
LG had not commented when Notebookcheck published its report. That leaves several factual questions unanswered, including what consent the test units had recorded, how long voice data remained on the set and which service received the upload. The company also needs to say whether the network discovery is required for a named feature or feeds advertising systems, and whether disabling that feature stops the traffic.
How to test a television you control
Owners do not need to assume that every LG set behaves identically. A network administrator can place a television on an isolated guest network or VLAN, record its model, region and firmware, then capture traffic at the router before and after changing ACR, advertising and voice settings. Repeating the standby, disconnect and reconnect sequence from the Gamers Nexus test can reveal whether the unit queues transmissions while offline. Traffic should be captured only on a network the tester owns or is authorized to inspect.
A useful record includes timestamps for power-state changes, DNS lookups, destination addresses and the exact consent screens accepted during setup. Encryption may hide the payload while still exposing when a set contacts a service and how much data moves. A repeatable trace will be more informative than a list of domains copied from another model because endpoints and firmware behavior can change; the original investigation likewise relied on Wireshark packet captures, according to Notebookcheck.
Gamers Nexus recommends disconnecting LG televisions from the internet and using an external streaming device. That is the clearest containment step in the published investigation, but it moves trust to another box rather than removing telemetry from the room. It may also disable updates and network features. Users who keep webOS online can at least isolate the set from personal computers, phones and storage devices, then check whether required streaming still works.
Network isolation cannot be treated as a microphone switch. In the reported test, voice input continued to accumulate while Ethernet was absent. Segmentation can limit what the television sees on the LAN, and blocking outbound access can prevent an immediate upload, but neither measure proves local capture has stopped. A verified software control or a physical loss of power provides a different kind of assurance.
The next useful evidence is concrete: an LG response, a list of affected models and firmware, results under each consent configuration, reproducible packet traces and a patch advisory for the webOS flaws after disclosure. Until those arrive, the measured conclusion is narrower than the viral one. Tested retail LG OLED sets retained voice input in standby and transmitted it after reconnection, while the same investigation observed local-network scanning. That is enough reason to treat a connected television as an active endpoint even when its screen is dark, without declaring every LG television compromised.