Anthropic Plans Invisible Watermarks for Claude-Generated Text and Files
TechTarget, Tuesday, August 19th, 2025
Anthropic will embed hidden machine-readable watermarks in Claude-generated text and images to meet EU transparency rules.
Anthropic is implementing invisible watermarks across all Claude products, APIs, and cloud services to identify AI-generated content. The company cited the EU's Code of Practice on Transparency as the primary driver, though it is applying watermarks worldwide in response to unreliable AI detection tools.
Watermarks will be embedded at the model level for text and use C2PA metadata standards for images. Watermarks indicate only that Claude processed content, not necessarily that it created it entirely, and their absence does not confirm human authorship.