In a completely unsurprising turn of events, Meta’s Muse AI is stealing Apple Messages, past and present, and uploading the contents to its cloud, even if explicitly told not to.

    • Leon@pawb.social
      link
      fedilink
      arrow-up
      18
      arrow-down
      4
      ·
      1 month ago

      Hard disagree! Yes, we know that Facebook does this. The average person doesn’t necessarily, and they still deserve privacy!

      This victim blaming nonsense plays right into the corpo playbook. It’s the privacy equivalent of “with what she was wearing she was practically asking for it” and we all know that’s bullshit.

  • tyler@programming.dev
    link
    fedilink
    arrow-up
    38
    arrow-down
    1
    ·
    1 month ago

    Just an FYI for anyone that didn’t know, your iMessages on your computer are literally just stored in a SQLite db. They’re very very easy to read without the app.

    • atomicbocks@sh.itjust.works
      link
      fedilink
      English
      arrow-up
      6
      arrow-down
      2
      ·
      1 month ago

      Source? AFAIK iMessages are stored encrypted by the Secure Enclave and can only be read through the same process that Touch ID works through. That’s why they can’t do iMessage through a browser because there must be a Secure Enclave present on the device. Pretty much the same reason Graphene OS only works on some devices.

      • tyler@programming.dev
        link
        fedilink
        arrow-up
        14
        arrow-down
        1
        ·
        1 month ago

        Me? I wrote an app to read my iMessage database to pull two factor codes out and type them in automatically cause I was lazy.

    • Pronell@lemmy.world
      link
      fedilink
      arrow-up
      3
      arrow-down
      1
      ·
      1 month ago

      Is Android any better in this regard?

      Not whatabouting, generally ignorant and curious.

      • tyler@programming.dev
        link
        fedilink
        arrow-up
        8
        ·
        1 month ago

        I don’t think Android has a corresponding computer operating system so no? This doesn’t apply to iPhones or iPads. Just Macs. And I’m not even sure if it applies to M series MacBooks, but it applied to the old intel ones about 5 years ago at least which was the last time I read my messages that way.

  • TootSweet@lemmy.worldM
    link
    fedilink
    English
    arrow-up
    9
    ·
    1 month ago

    without consent

    I’m surprised the EULA doesn’t say “by using Muse you agree to any fuckin’ thing Meta/Muse wants to do including ones we haven’t thought of yet”.

    Not that that constitutes “consent” in my book, but I wouldn’t think the article would say “without consent” if there was something in the EULA permitting them to do basically anything.

  • Rachel@lemmy.blahaj.zone
    link
    fedilink
    English
    arrow-up
    5
    ·
    1 month ago

    macOS is pretty protective over disk access as I have to constantly give apps permission to access folders like documents or iCloud. Though I assume anyone using something like muse is giving it full unrestricted access to the whole system.

    • jj4211@lemmy.world
      link
      fedilink
      arrow-up
      3
      ·
      1 month ago

      I don’t know the answer to this, but the AI agent would be at one of the “ends” of the conversation, so e2ee wouldn’t be a particularly relevant promise to that scenario.

    • GamingChairModel@lemmy.world
      link
      fedilink
      arrow-up
      1
      ·
      1 month ago

      The most secure messaging system in the world won’t matter if you read all your messages with a camera over your shoulder.

      E2EE protects what’s between the ends. But once a message is delivered to the end, the user can do whatever they want with it: screenshot or save the message, forward it to others, save it in a database open to others, etc.

  • jj4211@lemmy.world
    link
    fedilink
    arrow-up
    3
    ·
    1 month ago

    How many times does it have to happen for folks to understand “telling an AI not to do something” doesn’t work reliably? Feels like the AI providers need to explicitly direct them to provide an explanation that such directives aren’t reliable and any expression of an agreement/promise doesn’t mean what one word intuitively think.