Anthropic
Anthropic will quietly watermark all text and images Claude generates
🧒 Discovery
Claude will now slip a tiny invisible tag into everything it writes, like a secret fingerprint that says 'an AI made this'.
📰 Essential
To comply with the EU's AI rules, Anthropic is adding an invisible mark to Claude's text and provenance metadata to its images, a system that follows the content even after it's copied and pasted.
🔬 Expert
The move answers Article 50 of the EU AI Act, in force since August 2, which requires flagging AI-generated content. The text watermark stays invisible to readers but detectable with dedicated tools, while images carry the open C2PA provenance standard. One open question is how well these marks survive paraphrasing or screenshots, something regulators will likely watch closely.