August 13, 2026
Caught AI-handed
Claude users are mad that Anthropic's new watermarks will catch them using it
Claude’s secret labels spark panic, eye-rolls, and a whole lot of Reddit dunking
TLDR: Anthropic has started secretly labeling Claude-generated writing to meet European transparency rules, and users are split between calling it fair disclosure and treating it like a betrayal. The comments were harsher than the policy: lots of mockery, jokes about obvious AI prose, and predictions of a new watermark-removal arms race.
Anthropic thought it was doing the responsible thing: its chatbot Claude now adds an invisible label to writing it creates so computers can tell it came from AI. The move is meant to follow new European rules requiring AI-made content to be marked. But online? People instantly turned this into a full-blown comment-section soap opera.
The loudest critics acted like this was the end of civilization, warning that students, writers, and journalists would be “caught” using AI and stamped with a digital scarlet letter. That set off a second wave of reactions from people who were not remotely sympathetic. Some basically replied: if you’re pasting chatbot text into homework or news stories, maybe getting caught is the point. The mood swung hard between outrage and mockery, with one user telling the complainers to take “a deep breath,” while another roasted the whole meltdown with the now-iconic jab that someone “couldn’t even complain about Claude without using Claude to write it.”
Then came the smarter side-eye. Some users called the watermark idea hypocritical, saying it’s rich for AI companies to tag outputs after training on oceans of human-made writing. Others shrugged and said this is barely news because Google and OpenAI have already been doing similar things. And the funniest dunk of all? Several commenters argued Claude doesn’t need a hidden label because its writing is already so obvious that the “watermark” might as well be its vibe. In other words: the real story isn’t the policy — it’s the internet arguing over who’s embarrassed to be seen using AI.
Key Points
- •Anthropic added invisible watermarks to Claude-generated editorial text to mark it as AI-generated.
- •The article says the change was made to comply with the EU AI Act’s Transparency Code.
- •Some Reddit users argued the watermarking system could expose average Claude users while more sophisticated users could evade detection by paraphrasing outputs.
- •The article states that directly pasting AI-generated summaries or rewritten text into journalism or student essays would be unethical.
- •Despite criticism, the article reports that many users supported watermarking as a way to detect algorithmically generated material.