Coders Say They Already Found Workarounds to Claude’s Invisible Watermarks
TL;DR
Anthropic announced last week that it would embed invisible watermarks in AI-generated content to comply with new EU rules. Within hours, overrides were being touted online. How reliably those workarounds actually strip the marker has not been independently verified. The case shows how quickly technical labelling requirements come under pressure once they ship.
Nauti's Take
Watermarking is a sensible first step, and labelling AI output at least gives platforms and auditors a chance to trace where a text came from. The limit shows up immediately.
A marker that a round of light editing can strip will stop no serious misuse. Companies with disclosure duties still gain something here, while anyone treating it as fraud protection is planning too optimistically.
Sources