Anthropic is adding special markers to content created or processed by Claude. New Claude models will include these markers from the start, while Anthropic is also working on adding them to older models. The system is intended to provide more transparency about AI involvement without affecting the quality or readability of generated content.
Claude will use two main types of markers. Text will contain an invisible watermark that can stay with it after copying and some editing. Supported files will receive digital information showing that Claude processed them. This information will use the C2PA standard, which is designed to track where digital content comes from.
The markers will work across Claude, its API, Claude Code, Claude Cowork, and Claude Tag. They will also be available worldwide and on supported cloud services.
However, the markers will not provide definite proof that Claude created a piece of content from scratch. Someone may use Claude to edit, translate, or summarize material they created themselves. At the same time, heavy editing, translation, file conversion, or other changes can remove a marker. Content created with older models or unsupported features may also have no detectable marker.