I think the main reason is that in the future when training Calude on data obtained from the internet, they could discard first all data that were produced by claude in the first place.
this post was submitted on 20 Aug 2026
299 points (98.4% liked)
Microblog Memes
12050 readers
2250 users here now
A place to share screenshots of Microblog posts, whether from Mastodon, tumblr, ~~Twitter~~ X, KBin, Threads or elsewhere.
Created as an evolution of White People Twitter and other tweet-capture subreddits.
RULES:
- Your post must be a screen capture of a microblog-type post that includes the UI of the site it came from, preferably also including the avatar and username of the original poster. Including relevant comments made to the original post is encouraged.
- Your post, included comments, or your title/comment should include some kind of commentary or remark on the subject of the screen capture. Your title must include at least one word relevant to your post.
- You are encouraged to provide a link back to the source of your screen capture in the body of your post.
- Current politics and news are allowed, but discouraged. There MUST be some kind of human commentary/reaction included (either by the original poster or you). Just news articles or headlines will be deleted.
- Doctored posts/images and AI are allowed, but discouraged. You MUST indicate this in your post (even if you didn't originally know). If an image is found to be fabricated or edited in any way and it is not properly labeled, it will be deleted.
- Absolutely no NSFL content.
- Be nice. Don't take anything personally. Take political debates to the appropriate communities. Take personal disagreements & arguments to private messages.
- No advertising, brand promotion, or guerrilla marketing.
RELATED COMMUNITIES:
founded 3 years ago
MODERATORS
Slop Recursion Avoidance
That's pretty much certainly the reason, they don't want to train the next model on the output of the last one.
so you're saying add their watermark to literally everything created digitally to omit it from their garbage? cool
Why wouldn’t they?
Not exactly bulletproof, is it?
A successful detection result would indicate that Claude may have processed content. It would not prove that Claude was the original author, because users can submit human-created text, research, images, or files to Claude for summarizing, editing, translation, or proofreading.
Likewise, the absence of a watermark should not be treated as proof that a human wrote content. Older Claude models may not yet support marking, short text may contain too little material for reliable detection, and heavily edited output may lose its signal.
There are likely already open source tools that make a de-watermarked copy of the text
It says this works also for text copied from Claude and pasted into other applications. How is that possible? What degree of metadata storage is possible in text?
LLMs are big statistical networks that determine the following word in a sentence. If two words are just as likely to follow, Claude is gonna pick the next word based on a secret key and somehow this can be looked up later. That appears to be the gist of it if I understood it correctly.
Yep, also my understanding of the mechanism. For small texts it's probably not that hard to modify it enough to disguise it but the larger the text is, the harder it gets to remove the statistical watermark.
It's like a style, except it changes every word. If you know how, you can see when text is written in-style.
The mechanism happens in the way LLMs generate text sequentially. So as the LLM generates text, each time the LLM encounters a choice between two or more equally probable words and has to pick one, it chooses from a known pseudorandom key instead of its previous practice of actually relying on a source of randomness.
If you run the text through a checker script that tries to predict the next word, and takes note of every choice made in any given fork in the probabilistic flowchart, you can see that whole sequence of choices and compare to whether it matches a known key.
You'd need to know the model itself well enough to know when the text reaches a particular fork, and what the choices are in that fork, and the key itself, so even OpenAI/Google/Anthropic may not be able to run this watermark detection on their competitors' models. But they'd be able to know it themselves, for their own models.
Fontsize 0pt :)
I'm sure it has absolutely nothing to do with EU regulations that went into affect.
It very much seems to be a response to those regulations.
