Pinned
Frontier labs are baking invisible watermarks into generated text. The mark is a statistical bias in word choice: the model is nudged toward a secret, per-position "green list" of tokens, and a detector later flags text that leans green too hard.
No more! Change enough words
Claude's watermark probably doesn't work how you think. As the CTO of GPTZero, I'll explain how Anthropic, Google and OpenAI are building text watermarking in this brief explainer and whether it can be defeated.
Almost all forms of watermarking that are fast and cheap enough for






