Now Reading: Anthropic Adds Invisible Watermarks to Claude-Generated Text
-
01
Anthropic Adds Invisible Watermarks to Claude-Generated Text
Anthropic is adding invisible, machine-readable watermarks to text generated by supported Claude models, the company announced on Tuesday. The mark is designed to remain when text is copied and pasted, and may…
The Story in Brief
Anthropic is adding invisible, machine-readable watermarks to text generated by supported Claude models, the company announced on Tuesday. The mark is designed to remain when text is copied and pasted, and may survive some editing. New Claude models launched on or after 2 August 2026 will support the system from launch, while older models are being updated. The marking system is rolling out globally across Claude services, including its platform and cloud partners.
For supported files such as SVG, PNG and JPG, Claude will attach signed provenance metadata using the C2PA standard instead of a text watermark. Anthropic is still developing detection tools. It says failed detection will not prove that content was written by a human, since older models, extensive editing, paraphrasing or mixed material can remove or obscure the marker. The move follows the company’s commitment to the European Union’s AI Act transparency rules.
The Indian Opinion
Claims that watermarks will end AI-generated misinformation are overstated. Anthropic’s own limits show these markers are signals about provenance, not conclusive detectors. The opposite claim, that they are useless because editing can remove them, is also too sweeping. Their value will depend on accessible detection tools and adoption across platforms, not just Claude. The practical test is whether independent users can reliably identify marked content after ordinary editing.
Sources (2): gadgets360.com, timesnownews.com
This story was synthesised by AI from the 2 sources linked above.
Updated: this story now draws on 2 sources.
