Anthropic details Claude's text watermark: it only shows Claude was likely involved, is sparse in code and factual text, and disappears after a full rewrite (Anthropic)
Future Claude models will generate text that contains a watermark. This is a way of determining the likelihood that Claude …
Anthropic has revealed that Claude, their text-generating AI model, will now include a watermark. This watermark will indicate the probability that Claude was involved in creating the text. Major AI providers are implementing watermarks to comply with the EU AI Act. Anthropic explains that the watermarking method does not alter Claude's output quality, creativity, or readability.
The watermark is based on low-stakes random choices made during text generation, such as selecting between two plausible words. These choices create a pattern detectable only through a key, without influencing the meaning or experience of the reader. Claude's watermark is based on the SynthID-Text approach, which has been tested and found to have no impact on text quality.
However, the watermark can only determine the likelihood of Claude involvement, not confirm human authorship or identify other AI models.
Written by urgent.news from Techmeme'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.
- Anthropic certification: the Claude (CCAR-F) credential, explained dev.to
- Techies have concerns about Claude's hidden watermark. Anthropic has some answers. businessinsider.com
- European regulators seek Claude usage data from Anthropic, upsetting some users timesofindia.indiatimes.com
- Anthropic watermarking Claude AI content | Explained thehindu.com
