Link: Claude will apply invisible watermarks to AI text and images

Anthropic is set to embed machine-readable watermarks in Claude-generated text and images in response to the EU's AI Act. This regulation requires AI transparency and includes a grace period for compliance.

While the updates are not effective immediately, future Claude models will carry these identifiers from launch, with ongoing support planned for existing models.

The watermarks will be invisible but will help detect AI-generated content across various platforms globally, including various products like Claude Platform (API) and Claude Code.

For images, the C2PA standard will be used, already supported by major tech entities like Adobe and Google; however, details on the text watermarking process are less specific.

These embedded marks will travel with the content and remain even after editing, ensuring consistency regardless of where or how the content is used.

Anthropic aims to provide ways to detect these watermarks, although current tools like Google’s Gemini may not be compatible, showing the technological and implementation challenges that remain. #

--

Yoooo, this is a quick note on a link that made me go, WTF? Find all past links here.