Anthropic to Implement Watermarking and Metadata for Claude AI Content
5 sources across 4 countries · India · Argentina · Germany · Spain
Who reported this
- MediaNama
- The Indian Express
- Infobae
- Der Spiegel
- El Pais
What the colours mean
- Left
- Centre-left
- Centre
- Centre-right
- Right
- A hatched block means the outlet is affiliated with, or controlled by, a state.
Political lean describes where an outlet sits within the politics of its own country. It is never a position on a single global scale.
Political lean is comparable inside one country and not across them, which is why the bar groups by country first. Publicly funded broadcasters are not marked as state-linked.
The owner of each outlet is listed as a matter of record, not as a judgement about the outlet.
Anthropic is introducing invisible watermarks for text and provenance metadata for media generated by its AI chatbot, Claude. The company stated that these measures are being implemented to comply with the European Union AI Act, specifically the Code of Practice on Transparency of AI-Generated Content. The watermarking system is designed to be imperceptible and will not alter the meaning, quality, or readability of the text. According to Anthropic, the watermark is embedded at the model level and is intended to persist when text is copied and pasted or subjected to some editing. For media files such as PNG, JPG, and SVG, Anthropic will use the C2PA open standard to record provenance metadata, which allows users to detect if a file was processed by Claude or if it has been tampered with. These markings will be applied globally across all regions where Claude is available. Anthropic noted that models launched in the EU on or after August 2, 2026, will support machine readable marking at launch. The company cautioned that the system is not a definitive solution for determining authorship. A watermark may be present even if the underlying ideas came from another source, such as during translation or summarization. Conversely, the absence of a mark does not guarantee that text was not generated by AI, as marks can be removed through heavy paraphrasing, translation, or if the text passage is too short. Anthropic plans to release technical documentation and detection tools for third parties soon.
How each side framed it
- Centre-left
- Framed the move as a response to EU transparency regulations and highlighted the limitations of the technology.
- Centre
- Focused on the technical specifications and the comparative regulatory landscape between the EU and India.
- Centre-right
- Emphasized the global application of the measure and the industry standards involving other tech giants.
Sources
- Centre-left Der Spiegel: Claude: Chatbot from Anthropic marks AI texts of future models with watermarks
- Centre-left El Pais: It will no longer be so easy to copy and paste with AI: Claude will put a watermark on all texts it generates
- Centre-right Infobae: Claude will apply invisible watermarks to the text it generates to comply with the EU AI Act
- Centre MediaNama: Anthropic to embed watermark and C2PA metadata for AI-generated text and media by Claude
- Centre The Indian Express: Why Anthropic is watermarking text generated by Claude AI, how it will work
100% of the statements in this article were traced back to the source articles listed above.