How to know if content was made by AI: Claude adds invisible watermarks
Anthropic is introducing machine-readable marks for content generated by its Claude AI models as part of its commitments under the European Union’s AI Act, the company said, Qazinform News Agency correspondent reports.
Claude models launched in the EU on or after August 2, 2026, will support marking from launch. The system will apply to supported Claude models worldwide across Claude Platform, Claude, Claude Code, Claude Cowork and Claude Tag, although some platforms and features may not support every type of marking.
Anthropic uses two methods to mark content. For text, Claude embeds an imperceptible watermark directly into its output. The watermark does not affect the meaning, quality or readability of the text and can remain when the text is copied and pasted, as well as after some editing.
For supported files, including SVG, PNG and JPG formats, Claude adds digitally signed provenance metadata based on the Coalition for Content Provenance and Authenticity (C2PA) standard. The metadata can indicate that a file was processed by Claude and help determine whether it has been altered.
The company is also developing tools that will allow users and third parties to detect Claude’s watermarks and provenance metadata. Anthropic said it will publish technical details about the detection methods in future documentation.
The system has limitations. A detected mark shows that content may have been processed by Claude, but does not prove that Claude created the original material. For example, users may employ Claude to proofread, translate or summarize existing content.
At the same time, the absence of a detectable mark does not prove that content was not generated or processed by AI. Marks can be lost through heavy editing, paraphrasing, translation, file conversion or screenshots. Short passages may also contain too little text for reliable detection.
Anthropic is working to add marking support to models released before August 2, 2026, which are covered by a transition period under the EU AI Act. The company also advised developers using Claude in their own products to assess separately what transparency requirements apply to their services under Article 50.
Earlier, Qazinform News Agency reported that a new font was created to protect online writing from AI.