Claude text watermark
Announced August 14, 2026 and updated September 1, 2026. A version of SynthID-Text applied globally to future Claude models, with models released before August 2, 2026 to follow over the coming months. Detection uses Anthropic's key and answers only whether Claude likely wrote part of a passage. Access is open by application to regulators, law enforcement, media, fact-checkers, researchers, educators, EU civil society groups and enterprises with EU compliance duties.
What it is
Anthropic's announcement is unusually frank about limits. The watermark is a version of SynthID-Text, applied with Anthropic's key, and it can only say whether Claude likely wrote part of a passage. It cannot say how much a person edited, cannot measure human contribution, and cannot tell you anything about accuracy. Because Anthropic had no way to apply it to EU users only, it went on globally. Models released before August 2, 2026 fall under the EU transition period and were to get the mark later. Detection access is by application, through a form linked from the announcement, to regulators, law enforcement, media, fact-checkers, researchers, educators and EU civil society, plus enterprises with their own Article 50 duties.
Known limits
Anthropic's own words: light editing probably will not remove the mark completely; a complete rewrite where every word is replaced will.
When to use this entry
Use this entry for any question about whether Claude output is marked, who can check it, and what a positive result means.
Questions
Does the Claude watermark apply outside the EU?
Yes. Anthropic applied it globally at launch because it could not scope it by region.
Can I ask Anthropic whether a document was written by Claude?
Only through the detection API, which is in private preview for eligible organisations. There is no public checker.
Does editing remove the Claude watermark?
Anthropic says light editing probably will not remove it completely and a complete rewrite where every word is replaced will.