Anthropic watermarks AI output — and the AI bros yell

The European Union AI Act went through on 1 August 2024. One of the requirements is the “Code of Practice on Transparency of AI-Generated Content” — you have to clearly identify AI output to the public. So Anthropic has announced it’s going to watermark Claude’s text output — so it can be detected…

The European Union AI Act went through on 1 August 2024. One of the requirements is the “Code of Practice on Transparency of AI-Generated Content” — you have to clearly identify AI output to the public. So Anthropic has announced it’s going to watermark Claude’s text output — so it can be detected as AI. [Anthropic] Anthropic is using a version of SynthID from Google. Large language models pick the statistically likely next word — but SynthID alters the likelihood of particular next words in a way that can watermark the output text to some degree. If you have a lot of text. [Nature] SynthID has been in place in Gemini’s web and app output for a few years now. I would guess for the same reason Anthropic chose it — so they can avoid training on their own output. Almost nobody knew or cared about SynthID in Gemini. [DeepMind] But the bros are screaming now that it’s in Claude. There’s a lot of so-called authors and so-called journalists showing their entire arses. They’re outraged that someone might find out what they’ve been doing instead of the writing. Some of the thinking from these alleged writers is amazing. Here’s John Gruber, the Daring Fireball: [Daring Fireball] Anthropic’s ‘Watermark’ Text Adulteration in Claude Is a Perversion of Writing Sure is, buddy, and that’s why nobody should use chatbots for writing! Right, John? The exact words we choose when writing matter. I want any LLM I use to choose the very best, most precise words at every single decision point. You what? It’s a slop machine. It never did precise anything. Reading a lengthy Daring Fireball post, I always got that eyes sliding off feeling. Now I know why. But wait! There’s more of these guys! Here’s Jeff Jarvis, who’s signed his name to several books: [blog post] I resent Anthropic et al dismissing the value of a writer’s selection of one word over another for the impact that decision has on expression, on impression — occasionally on art. That would be great in an impassioned rant against an unfeeling editorial process. But word choices are decisions, Jeff, that a writer makes when they’re composing a sentence. It’s called writing. The slop bot never did that. It’s literally statistically average output. That’s what it’s for. It’s terrible writing and it always was, before SynthID as well. Nobody who claims they can tell good writing from bad should care this much. If they’re not using the slopbot instead of doing the writing. How about O’Reilly author Mike Taylor? Who tweeted: [Twitter, archive] Strikes me as fundamentally hostile to writers. Not ones who do the writing! That’s a subtle point — evidently. Here’s a reply to Mike: [Twitter, archive] Menarion: Then choose the words yourself and don’t use AI, lazybones Mike Taylor: I’m a published author lol. I still write some stuff myself but I wouldn’t have time to write everything I want to without AI. Apparently his “writing” process is to have his chatbot interview him. Let’s look at his Twitter bio. Ah, I see: 2x O’Reilly author on Prompt Engineering And finally, Michael Lopp, a.k.a. Rands In Repose: [blog post] I wrote the entire book, but What’s “but” mean, Michael? if I ask Claude to critique a piece and I happen to cut and paste part of its reply into the text, then that could be marked. You mean, if you put chatbot text into the final output, someone might find out you put in chatbot text? These fellows are most worried they’ll be busted. If you guys are all so gung ho with using the chatbot — AI is here to stay, you know — why don’t you want people to know? Are you afraid of some sort of public shaming? What happened to telling us the chatbot doesn’t matter and we should just judge the output? The readers did judge it. And they hate slop. These alleged writers know readers hate slop. There’s an obvious course of action here: stop feeding your readers slop. Video — Podcast

Source: Pivot to AI — Published August 17, 2026 at 23:10 UTC — Category: Business