Anthropic's Claude now embeds an invisible "Scarlet Letter" watermark on any text it processes, even when that text is later edited by a human. The watermark currently remains hidden but will flag content as originating from the AI model. This approach enables detection of AI‑generated text