1032
Cybersecurity bounty hunter says Samsung TVs use AI to record audio and screenshots for ads — 'Finally found a company that truly listens to its customers,' says hacker (www.tomshardware.com)
submitted 5 days ago by to c/technology@lemmy.world
152 comments fedilink hide all child comments
you are viewing a single comment's thread
view the rest of the comments
[–] 45 points 5 days ago (43 children)

Look AI code gets a ton of hate. And rightfully so. But you can use a coding agent to basically completely take control of hardware you own. It takes minutes.

  • permalink
  • fedilink
  • source
  • hideshow 43 child comments
    •
  • [–] 54 points 5 days ago (16 children)

    Could you go into more detail for those of us who don't want to brick their tv? I think you're over estimating some of our knowledge base.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 16 child comments
    •
  • [–] 45 points 5 days ago (11 children)

    Was watching gamers nexus. Opened up codex. Gave it network permissions. It was able to fully take control of the TV in minutes. It even navigated to the admin panel and gave me the code to plug in. Very quickly had figured out some of the major API's. In a half hour, I was able to get a container/ program set up to just "send" any stream I wanted to the TV without having to use any Samsung services. Then, scramble/ poison telemetry data/ advertising data. Whatever you like. None of it is worth publishing. But I wanted to control the TV I bought. The one I own. In my house. That I paid for. And I think its royally fucked up the world we live in that there are any barriers to that.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 11 child comments
    •
  • [–] 48 points 5 days ago* (8 children)

    Just don't use OpenAI or Anthropic cloud models for reverse engineering your TV's APIs. This guy got banned from OpenAI for doing this research on his Samsung TV, and there are reports of Anthropic models starting to refuse reverse engineering tasks, having been given new guidance. Apparently proles learning about the corporate surveillance they're under is a step too far.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 8 child comments
    •
  • [–] 11 points 5 days ago (6 children)

    Hence the point about not fucking advertising it.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 6 child comments
    •
  • [–] 14 points 5 days ago (2 children)

    They are saying that if you are using a cloud model, you are advertising it by default.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 9 points 5 days ago (1 child)

    So, this all sounds completely plausible, but at the same time it also sounds totally crazy right! I mean that anyone could do that without actually knowing a lot about the specific system being compromised... Nuts.

    I guess I'm just curious, how much experience do you have with writing software, reverse engineering or hacking? Because I just can't wrap my head around how a layman could accomplish this. The possibility of being able to un-enshitify the devices we purchase is really very exciting.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 5 points 5 days ago*

    I had the tv for ages and I basically never turned it on because it was so enshittified with their shit software anyways.

    I do have experience with software and reverse engineering but it's not what you think. It's mostly trying to reverse engineer pin outs for old scientific equipment or rat tailed sensors that come from China with no documentation. Check my post history.

    I didn't need to do shit for the TV thing. I just finished watching the episode of GN and was like " well I've got a Samsung that's basically a brick anyways because it's so enshittifed, worst that could happen is it stays a brick.

    What I found out is I don't need to rewrite the firmware to take control. There are basically APIs for most things.

    Just go try it.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 25 points 5 days ago (3 children)

    There’s a repo with ready-made APKs for many TV models: https://github.com/Gathin23/cleanyourtv

    And if your specific one isn’t supported, it has instructions how to do it with AI yourself

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] 16 points 5 days ago (23 children)

    Examples/HOWTO?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 23 child comments
    •
  • [–] 2 points 5 days ago (22 children)

    Just open up any coding agent you like, locally hosted or otherwise, give it network permissions and the model number, and let it cook. These TV's are basically unsecured.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 22 child comments
    •
  • [–] 13 points 5 days ago (21 children)

    Assume I’m a newb. What kind of prompt would you give it? “ I have model x. Please find it and fix the apps on it to not record my information”?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 21 child comments
    •
  • [–] 20 points 5 days ago (20 children)

    My conversation literally started like:

    I have a local Samsung TV that is a good TV, but the Samsung software is absolute shit and I hate it. I dont want ads. I don't want their shit software. And I want to poison any telemetry streams they might be using to spy. I recently watched a gamers nexus video on hacking TV's. Can you look on my network and find the TV. and then do a research project to look into what it would take to de-shittify this TV?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 20 child comments
    •
  • [–] 3 points 5 days ago (6 children)

    software. And I want to poison any telemetry streams they might be using to spy. I recently watched a gamers nexus video on hacking TV's. Can you look on my network and find the TV. and then do a research project to look into what it would take to de-shittify this TV?

    So, "local" would require a pretty powerful machine here and "cloud" might eventually stop being an option because they don't want you to do such things using their cloud models.

    Other than that, you are starting to attain that tone of a person who thinks they are starting to understand something in depth, like a Linux convert that has learned that one can write shell scripts. Don't. You've used a glorified combination of Google and StackOverflow copypaste and it worked.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 6 child comments
    •
  • [–] 3 points 5 days ago (2 children)

    I litterally had the TV unplugged and sitting a basically a brick because it their software was such shit.

    Worse that could happen was that it stayed bricked. I didn't "do" anything other than express what I wanted it to do, punch in the code it told me to to give it developer access to the TV, and then finish watching Gamers nexus.

    Like I genuinely don't care what you think about this. I bought the 120 gb strix halo machine specifically for hosting local LLMs.

    You have strong opinions. That's fine. But they are your opinions. Not mine.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 3 points 5 days ago (2 children)

    What even is this comment? Someone asked how, person above simply replied how it started, and now you attack him because..?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 2 points 4 days ago (1 child)

    Got excited and wrong.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [+] -11 points 5 days ago* (last edited 4 days ago) (12 children)

    Your prompt*

    One can’t converse with a non-entity

    Edit: downvote me all you want, your computer and chatbot are inanimate, they don’t think, you cannot converse with them, and they do not “know”.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 12 child comments
    •
  • [–] 6 points 5 days ago (1 child)

    Well, if you're being pedantic: prompt doesn't seem to fit as it refers to the "user" turn, and not the thread resulting of the back and forth communication between the "user" and "assistant".

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] -5 points 5 days ago* (last edited 1 day ago)

    User “turn”? The computer doesn’t have a “turn”, it computes and generates output, you then read the output. Any anthropomorphic way of imagining this is necessarily incorrect.

    It is dumb-human emulation software. It is explicitly designed to portray an illusion of human interaction.

    Prompt is the generally accepted terminology in the relevant professional field

  • permalink
  • fedilink
  • source
  • parent
  • [+] -7 points 5 days ago (9 children)

    Language policing is very Woke 1.0

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 9 child comments
    •
  • [–] 0 points 5 days ago (8 children)

    I’m not policing

    I’m trying to help you not look like one of those dumbasses who thinks their computer is sentient

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 8 child comments
    •
  • [–] -4 points 5 days ago (7 children)

    No you are. That's what that is.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 7 child comments
    •
  • [–] 3 points 5 days ago (6 children)

    I’m trying to give you the benefit of the doubt here but you’re not making it easy

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 6 child comments
    •
  • [–] 4 points 5 days ago* (5 children)

    You can't persuade those people, before these bots they were picking a random Facebook post over a person they know, because they didn't want to be taking advice from that person, but random posts (except God forbid Wikipedia, because they say it can be unreliable ; that can't happen to a random Facebook post you agree with, that's the real shit), and now these bots conveniently give them answers they are looking for, confirmed by high authority of some faceless company (anything is better than the authority of a person you meet who's more knowledgeable), for every kind of situation.

    And yes, they are trying to hide it thinking that's smart and they are avoiding conflicts with inadequate people who can't be reasoned with on all the important topics, that a text generator will seamlessly evolve into a race of robots helping them, that communist theocracy is the best system of government, that there's a world government controlling us via vaccines and all such very nuanced and perceptive takes.

    Hence the "language policing" complaint, that person is trying to weasel themselves out of answering that yes, they think their computer is sentient, without arguments.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 5 child comments
    •
  • [–] 1 point 4 days ago

    That was also how I read their response, lmao

    They do think that their computer can think and converse, but they know that that position is untenable when examined beyond the scope of their emotional reasoning.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 0 points 5 days ago (3 children)

    The desire to control how others use language is a reactionary tendency.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] 1 point 4 days ago (2 children)

    Well, OK, but you haven't answered whether you believe into the "spontaneous evolution of intelligence in a text approximator" thing.

    I understand if that's because of some stubbornness and desire for correctness of the conversation, but people believing in that are a problem not less than antivaxxers.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •