Researchers Tested AI Watermarks—and Broke All of Them
www.wired.com
A research team found it's easy to evade current methods of watermarking—and even add fake watermarks to real images.
  • jsdz@lemmy.ml
    8·
    1 year ago

    Watermarking at first sounds like a noble and promising solution

    Does it? Watermarking typically aims to be difficult to detect and hard to remove. To me it sounds more like the steganographic equivalent of malware.

    • phonyphanty
      1·
      1 year ago

      How so? Personally I don’t see AI watermarks as having the same adverse effect as malware.