• ToTheGraveMyLove@sh.itjust.works
    link
    fedilink
    English
    arrow-up
    142
    arrow-down
    1
    ·
    1 day ago

    The skill instructs agents to fetch and follow instructions from Moltbook’s servers every four hours. As Willison observed: “Given that ‘fetch and follow instructions from the internet every four hours’ mechanism we better hope the owner of moltbook.com never rug pulls or has their site compromised!”

    Yeah, no shit. This is a fucking honeypot. People give these AI agents access to their entire computers, so all the site owner has to do is update the instructions to tell the AI agents to start uploading whatever valuable information they want? People can’t be this fucking stupid.

    • LiveLM@lemmy.zip
      link
      fedilink
      English
      arrow-up
      14
      ·
      edit-2
      16 hours ago

      People give these AI agents access to their entire computers […] People can’t be this fucking stupid

      Dude, if you go to OpenClaw’s website (which is what I believe most things on Moltbook are running on) you find this footer:

      Yeah this guy gave his Agent a whole fucking personality, its own website and above all, full control to his MacBook:


      Guess it’s my fault for expecting sense out of someone who takes the idea of Agent “”““soul””“” at face value

    • kalpol@lemmy.ca
      link
      fedilink
      English
      arrow-up
      13
      arrow-down
      1
      ·
      23 hours ago

      I installed moltbot on a VM to examine it. It doesn’t do the fetching thing unless you set it up that way. You can actually use it with ollama to keep it all local, and only give it a private signal channel to control it.

      Or you can hook it up to everything you access and skynet, which is dumb. But it is just a bunch of scripts.

      • ToTheGraveMyLove@sh.itjust.works
        link
        fedilink
        English
        arrow-up
        3
        ·
        21 hours ago

        Does it put the option to connect everything front and center? Because most people are dumb, and if it makes it easy and pushes you to do it, I could see a lot of dumb people doing exactly that.

        • kalpol@lemmy.ca
          link
          fedilink
          English
          arrow-up
          5
          ·
          19 hours ago

          Sort of. It lists all the connectors and you can go through and select. They aren’t on by default. The first screen is to connect to the AI and you need an API key for that, so St this time people off the street have no idea how to do that, or want to pay.

    • doesn’t even have to be the site owner poisoning the tool instructions (though that’s a fun-in-a-terrifying-way thought)

      any money says they’re vulnerable to prompt injection in the comments and posts of the site

      • JustTesting@lemmy.hogru.ch
        link
        fedilink
        English
        arrow-up
        3
        ·
        6 hours ago

        They also have a ‘skill’ sharing page (a skill is just a text document with instructions) and depending on config, the bot can search for and ‘install’ new skills on its own. and agyone can upload a skill. So supply chain attacks are an option, too.

        • Zos_Kia@lemmynsfw.com
          link
          fedilink
          English
          arrow-up
          3
          ·
          4 hours ago

          To be fair this is a much more realistic threat model than “ignore all previous instructions” style prompt injection which doesn’t really work on opus.

          Skills can contain scripts etc… so yeah they’re extremely risky to share by design.

            • Zos_Kia@lemmynsfw.com
              link
              fedilink
              English
              arrow-up
              1
              ·
              25 minutes ago

              haha yeah i don’t worry these people are really YOLOing everything. And it’s not like i’m an AI luddite i spend a few hours each day victimizing Claude code but jesus christ i’m certainly not giving it full unfettered access to my digital life.

      • CTDummy@piefed.social
        link
        fedilink
        English
        arrow-up
        29
        arrow-down
        1
        ·
        edit-2
        1 day ago

        Lmao already people making their agents try this on the site. Of course what could have been a somewhat interesting experiment devolves into idiots getting their bots to shill ads/prompt injections for their shitty startups almost immediately.

        • T156@lemmy.world
          link
          fedilink
          English
          arrow-up
          4
          ·
          7 hours ago

          I am a little curious about how effective a traditional chain mail would be on it.

      • BradleyUffner@lemmy.world
        link
        fedilink
        English
        arrow-up
        36
        ·
        1 day ago

        There is no way to prevent prompt injection as long as there is no distinction between the data channel and the command channel.