Anthropic to watermark Claude-generated text, images

Anthropic's move to watermark Claude-generated content reflects growing industry concern over the origin and authenticity of AI-produced material. As Claude is deployed across a wide range of applications — from customer-facing chatbots to content creation platforms — the ability to cryptographically verify that output came from the model addresses a practical need: distinguishing human-authored from machine-generated text and images. The announcement came via a help desk article rather than a formal press release, suggesting the feature may have been implemented or is being rolled out gradually. The invisible nature of the watermark is strategically important, as visible markers would degrade user experience and could be stripped by bad actors; by contrast, embedded metadata or steganographic markers are harder to remove without degrading the underlying content.