How Claude's Text Watermark Works
Anthropic explained the text watermark coming to future Claude models, added to meet EU AI Act requirements. When several word choices are equally valid, t
Claude shipped a new update: How Claude's Text Watermark Works.
Anthropic explained the text watermark coming to future Claude models, added to meet EU AI Act requirements. When several word choices are equally valid, the model biases that pick using a hidden pattern that Anthropic can later detect, so Claude's involvement can be verified without changing output quality or spending extra tokens.
Claude
AI assistant by Anthropic — write, analyze, code, and create with one of the most capable large language models available.
View Claude →