Anthropic has announced a new machine-readable watermarking system for content generated or processed by its Claude AI models, adding invisible labels to text and provenance metadata to supported files. The move is aimed at improving transparency around AI-generated content and aligns with requirements under Article 50 of the European Union's AI Act.
Invisible Watermarks For Text
Under the new system, supported Claude models will embed an imperceptible, machine-readable watermark directly into generated text. Anthropic said the watermark cannot be seen by humans and does not affect the meaning, readability or quality of the output.
ALSO READ: Nitin Gadkari's Stark Warning On Road Safety: 20 People Die Every Hour, No War Claims As Many Lives
The watermark can remain intact when text is copied and pasted and may survive some forms of editing. However, extensive editing, paraphrasing, translation or combining text from different sources can make the watermark undetectable, as per reports.
Provenance Data For Files
For supported file formats, including PNG, JPG and SVG, Claude will add digitally signed provenance metadata based on the C2PA open standard. The metadata can indicate that a file was processed by Claude and help determine whether it has subsequently been tampered with.
Anthropic is also developing tools that will allow users and third parties to detect the embedded watermarks and provenance information.
Not Proof Of AI Authorship
Anthropic stressed that detecting a watermark does not necessarily prove Claude originally created the content. A user could have employed Claude only for proofreading, translation, summarisation or file conversion.
Similarly, the absence of a watermark does not prove that content was not generated using AI, particularly if it has undergone extensive editing or comes from an older model that does not support the system.
Rollout Across Claude Ecosystem
The marking system will apply at the model level across Claude, Claude Platform, Claude Code, Claude Cowork and Claude Tag, as well as services provided through AWS, Google Cloud and Microsoft Foundry, India Today reported.
ALSO READ: Gold ETFs See 55% MoM Fall In July Inflows: AMFI Data
Anthropic said all new Claude models will include the technology from launch, while models released before August 2, 2026, are being updated to support the system. The approach is comparable to Google DeepMind's SynthID and comes amid growing efforts to identify AI-generated content.
Essential Business Intelligence, Sharp Market Insights, Practical Personal Finance Advice, Daily Fuel, Gold and Silver Prices and Latest Stories — On NDTV Profit.
