On 11 August, Anthropic announced that all future Claude models will generate text with a watermark indicating its artificial intelligence origin, according to IEEE Spectrum AI.
Anthropic is not the only company using this approach: Google applies its own text watermark to Gemini model outputs, and Anthropic's solution is based on Google's development. OpenAI has not yet implemented a text watermark but plans to do so.
For users and editors, this means the potential emergence of an additional signal regarding text origin. However, available materials do not explain exactly how the marking will work or how easily it can be detected or preserved during editing.
editorial commentary
Why it matters
A likely consequence is increased attention to the origin of machine-generated text among editors, users, and model developers. The nearest observable signal will be the publication of technical details or the appearance of markings in new Claude models. Significant uncertainty remains regarding the watermarking method, implementation timelines, and its robustness after editing.