• LostCarcosan@lemmy.today
    link
    fedilink
    English
    arrow-up
    2
    ·
    1 day ago

    I guess that’s a good argument about why it should be mandatory that all AI generated images should have some kind of watermarking to make them identifiable

    • 9bananas@feddit.org
      link
      fedilink
      English
      arrow-up
      1
      ·
      1 day ago

      generally a good idea, but not for this:

      CSAM distributors will quickly figure out how to add the watermark to real content; what then?

      then you’ve got an even bigger problem on your hands…

      • jumping redditor [they/them]@sh.itjust.works
        link
        fedilink
        English
        arrow-up
        0
        ·
        12 hours ago

        that could be mitigated with cryptographic signatures being attached to the direct output from corporate models and either needing that signature to match or having the full prompt and random numbers for a reproducible output.