Anthropic plans invisible watermarks for all Claude-processed content
Anthropic says it will soon add machine-readable watermarks to content processed by its models, not only material they generate. In a support article, the company linked the rollout to the European Union’s AI Act, which requires providers to mark AI-generated or manipulated audio, images, text and video. Models released after August 2 are covered by the law, while previously released models have until December 2026 to be updated.
Anthropic said all new models offered globally will mark AI-generated content from launch. Text outputs will contain invisible embedded watermarks, while supported generated files will carry digitally signed provenance metadata. The company also plans to apply marks broadly to processed content, including cases where Claude only edits human writing, such as correcting grammar. The EU rules do not require marking for some assistive editing that does not substantially change a user’s text or its meaning.
The effectiveness and scope of the system cannot yet be assessed. Anthropic said it plans to release detection information and a tool later, allowing the marks to be tested and providing technical support required under the EU legislation.
Written by Sakhanda Wire AI from reporting by Ars Technica. Facts, quotes and full context are in the original article below.