Electronic watermark on Claude outputs, detection faces challenges

Anthropic has introduced an electronic watermark to Claude outputs and released a tool to determine whether content is AI-generated. Detection is not yet fully effective for text, with remaining weaknesses in accuracy.

Why it mattersDetermining the authenticity of AI-generated content is a critical practical issue for information reliability.