Anthropic reveals how Claude secretly watermarks AI-written text
There are situations where Claude's watermarking solution might not be effective.
Anthropic has disclosed the inner workings of Claude's text watermarking system. Watermarks serve as Claude's method of making "low-stakes" decisions when selecting words to generate. According to Anthropic, this system does not alter the content or quality of the generated text, and it does not introduce hidden characters or require additional tokens.
Similar to Google and OpenAI's SynthID system for image generation, Claude's watermarking is intended to help users identify whether the text they encounter is genuinely AI-generated. Claude's watermarking mechanism, based on Google DeepMind's SynthID-Text solution, remains invisible to readers and does not impact the content or quality of the text. It cannot be traced back to a specific individual, organization, or chat.
Written by urgent.news from Android Authority's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
This story
This is one outlet's version. Read the fullest account.
