After AI company Anthropic introduced an “AI watermark” into text generated by its chat assistant Claude — one that’s detectable through quirks in writing style — some users began canceling their subscriptions over it. On 5ch, technical debate continued over how this invisible watermark actually works and how it differs from watermarks embedded in images.
Anthropic’s move to introduce a “watermark” into Claude-generated text is driving some users away from the AI assistant.
“It just keeps inserting watermarks no matter what.”
Richard Echols, an AI consultant based in Georgia, USA, revealed that he canceled his Claude Max subscription on the 12th, saying the watermark was a major reason behind the decision.
Source: businessinsider.jp / Original article here
Is it just geniuses and total idiots with Americans? No middle ground?
Re: #4
The genius and the idiot are the same guy
The genius and the idiot are the same guy
Re: #4
Europe used to treat Galileo like an idiot too, back in the day
Europe used to treat Galileo like an idiot too, back in the day
I always thought "cloud" was kind of a dated tech buzzword — now there's an AI service running around with basically the same name?
Re: #11
That's "Cloud." Claude is pronounced "Claude" (kurōdo, like the French name)
That's "Cloud." Claude is pronounced "Claude" (kurōdo, like the French name)
Re: #11
"Kurōdo" is also the word for a sake-brewery craftsman (a pun on how Claude is pronounced in Japanese)
"Kurōdo" is also the word for a sake-brewery craftsman (a pun on how Claude is pronounced in Japanese)
Re: #11
Stick "Maki" in front of Claude and the reading becomes obvious (a nod to Japanese actor Maki Kurando)
Stick "Maki" in front of Claude and the reading becomes obvious (a nod to Japanese actor Maki Kurando)
Can't you just strip the watermark out? Are people not even reading it and just posting?
Re: #14
It's expressed through quirks in sentence style and biases in word choice — it's not a watermark you can spot at a glance
It's expressed through quirks in sentence style and biases in word choice — it's not a watermark you can spot at a glance
Re: #18
Ah, I see.
> According to Anthropic, this watermark travels with the content even through copy-paste, and may survive some amount of editing.
Ah, I see.
> According to Anthropic, this watermark travels with the content even through copy-paste, and may survive some amount of editing.
Re: #18
This is presumably built for English, French, and the like. How would it even work for Japanese — some kind of vertical acrostic reading?
This is presumably built for English, French, and the like. How would it even work for Japanese — some kind of vertical acrostic reading?
Automatic "AI watermark" embedding will probably end up mandatory for all AI eventually. We can barely tell AI content apart from real stuff anymore, and fake AI campaign videos of rival candidates are already possible during elections. Digital watermarking is going to become mandatory worldwide.
Re: #51
Anyone who skips it, or illegally strips it off, will probably end up facing penalties
Anyone who skips it, or illegally strips it off, will probably end up facing penalties
Re: #51
Interesting angle — I bet pirated copies and government/military use will just quietly run unregulated versions
Interesting angle — I bet pirated copies and government/military use will just quietly run unregulated versions
If it's a document, just have it output as Markdown and open that in Word — there's no way for a watermark to survive that
Re: #116
It's using the fluctuations unique to how AI generates text — the writing itself is the basis for detection
It's using the fluctuations unique to how AI generates text — the writing itself is the basis for detection
Re: #116
It's not a binary-level watermark being inserted — the point is that quirks in the writing let you identify a specific author. Like if someone always ends sentences with "da'cha" or "dattebayo," you can tell who wrote it (catchphrases from the characters Lum and Naruto). The idea is to embed something like that naturally. Some people have a distinctive way of placing punctuation too, right?
It's not a binary-level watermark being inserted — the point is that quirks in the writing let you identify a specific author. Like if someone always ends sentences with "da'cha" or "dattebayo," you can tell who wrote it (catchphrases from the characters Lum and Naruto). The idea is to embed something like that naturally. Some people have a distinctive way of placing punctuation too, right?
It's not about sentence endings lmaooo. I already said it's the fluctuation in the text generation lol. Same mechanism as AI-detection tools.
Re: #126
Strictly speaking it's not just sentence endings, but what happens if someone builds a generator that impersonates a specific person? Like if you ask it to "write this in Hiroyuki's style" (referring to 2channel founder Hiroyuki Nishimura's distinct blunt tone), that would clash with the AI watermark. I'd guess building a "Murakami Haruki-style generator" or a "Hiroyuki generator" would conflict with the watermark lol
Strictly speaking it's not just sentence endings, but what happens if someone builds a generator that impersonates a specific person? Like if you ask it to "write this in Hiroyuki's style" (referring to 2channel founder Hiroyuki Nishimura's distinct blunt tone), that would clash with the AI watermark. I'd guess building a "Murakami Haruki-style generator" or a "Hiroyuki generator" would conflict with the watermark lol
Re: #133
That's just because you're assuming the watermark can only live in sentence endings lol
That's just because you're assuming the watermark can only live in sentence endings lol
Re: #133
Makes me curious how the AI's telltale fluctuations would get embedded when generating text in Dazai Osamu's style lol
Makes me curious how the AI's telltale fluctuations would get embedded when generating text in Dazai Osamu's style lol
For images, just saving as .jpg and changing the compression ratio seems like it'd wreck the watermark. They might design it to survive compression, but there are still plenty of ways to deliberately destroy it.
Re: #127
Whether you turn it into an image, compress it, or whatever, I think it's the same as long as it's still readable as text — the quirk is baked into the word choices themselves
Whether you turn it into an image, compress it, or whatever, I think it's the same as long as it's still readable as text — the quirk is baked into the word choices themselves
Re: #145
Oh, sorry, were you talking about images?
Oh, sorry, were you talking about images?
Re: #145
That's a slightly different topic. Text watermarks and image watermarks work on completely different principles. A text watermark is about stylistic quirks, but an image watermark operates at the binary level.
Look up "steganography" and you'll see — there have long been ways to embed extra data inside image files.
That's a slightly different topic. Text watermarks and image watermarks work on completely different principles. A text watermark is about stylistic quirks, but an image watermark operates at the binary level.
Look up "steganography" and you'll see — there have long been ways to embed extra data inside image files.
ChatGPT writes pretty distinctive text, doesn't it. I always wonder what's with the emoji it sticks in headings.
Re: #182
Huh, I don't get that at all
Huh, I don't get that at all
Re: #182
When it does add them, just tell it not to and they stop — same with other models
When it does add them, just tell it not to and they stop — same with other models
So what does a text watermark actually look like?
Re: #211
"Hello, my name is Yamada Taro. Nice to meet you."
→ "Yo, I'm Yamada Taro! Stoked to meet ya!"
Think of it as the writing style shifting like that (the second line mimics Son Goku's rough speech pattern from Dragon Ball)
"Hello, my name is Yamada Taro. Nice to meet you."
→ "Yo, I'm Yamada Taro! Stoked to meet ya!"
Think of it as the writing style shifting like that (the second line mimics Son Goku's rough speech pattern from Dragon Ball)
*This article is excerpted and summarized from the 5ch (News Express+) thread「🤖Users Begin Cancelling Claude Subscriptions… Backlash Against Automatic “AI Watermark” Feature: “It’s Disrupting My Work”」.
