• frongt@lemmy.zip
    link
    fedilink
    English
    arrow-up
    3
    arrow-down
    3
    ·
    10 hours ago

    It’s not sending it back. It’s essentially watermarking the output, so that it can be identified if it appears somewhere unexpected.

    • GamingChairModel@lemmy.world
      link
      fedilink
      English
      arrow-up
      15
      ·
      7 hours ago

      It modifies the prompt, aka the input, not the output. It is smuggling 3 bits of secret user/session data in a wrapper that doesn’t look like it contains that data. As the article explains:

      So the marker becomes part of the system context sent to the model.

      This is a normal timestamp on a prompt:

      Today's date is 2026-07-11.

      But if your system timezone is a Chinese mainland timezone, it looks like:

      Today's date is 2026/07/11.

      Then, if your base URL includes a keyword like “deepseek,” it silently replaces the apostrophe from a ' to a ʼ:

      Todayʼs date is 2026-07-11.

      Or if the base URL has one of the domains on the list, like any .cn domain, it replaces the apostrophe with another apostrophe character:

      Today’s date is 2026-07-11.

      And if it has both a URL and a keyword on the watchlist, the prompt context includes:

      Todayʹs date is 2026-07-11

      That’s 3 bits of information: does this system have a mainland Chinese time zone, does the base URL contain a known keyword (associated with Chinese AI competitors) or a known domain (associated with mainland China or its major tech companies). And it sneaks it on by without making it obvious.

      That’s steganography.

    • QuizzaciousOtter@lemmy.dbzer0.com
      link
      fedilink
      English
      arrow-up
      6
      ·
      9 hours ago

      Did you read the post?

      the marker becomes part of the system context sent to the model. (Where Anthropic probably parses in their backend)