Hold on - this is not a mark that gets stamped. It is a mark that gets rolled like dice.
Fact
The facts: on 14 August 2026 Anthropic published an explanation of the text watermark that future Claude models will carry. When the model chooses between equally suitable words, the choice is settled not by a random generator but by a key and the few preceding words; whoever holds the key can estimate the probability that a text came from Claude. Anthropic claims: no impact on quality, indistinguishable to the reader, nothing added to the characters, zero extra tokens or cost, nothing that leads to a specific person, organisation or chat. The method is Google DeepMind's SynthID-Text. The trigger is the EU AI Act - since 2 August 2026 providers on the European market must mark AI content; the other major developers signed the same Code of Practice and will have their own marks.
Let's be honest about what it is for. Not to catch the student - there are cheaper ways for that. It is so that someone with the key can say 'this is probably from a model'. Probably. Not certainly. Anthropic itself writes that the signal weakens in text a human has edited.
The mark says 'probably'. The law wants 'yes'.
We are a publisher. We already say when a machine has touched a visual. Text under our editorial responsibility is ours and carries our name - and no watermark changes that rule. But it is worth knowing: from here on, anyone who copies Claude output one to one carries the mark with them.
Read their explanation. It is short and honest about its limits - a rare quality on this subject.
Official primary sources
Anthropic - How Claude's text watermark worksThe visual is generated code-art, no third-party images.