Anthropic watermarks Claude outputs for EU AI Act — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Anthropic will add machine-readable watermarks to Claude-generated text and files in its new models, per the source
- The watermarking system is designed to comply with new European Union transparency regulations, specifically the EU AI Act
- The watermarking covers both text and image outputs, effective August 2 per prior coverage of the same rollout
- Communications teams using Claude for routine tasks — cleanup, translation, or formatting of human-drafted press releases — could stamp those documents with machine-readable AI markers, according to the source's "Why it matters" section
Why it matters: Anthropic's watermark fires indiscriminately: every Claude-generated file gets tagged, including human-drafted documents that pass through Claude for routine editing. PR and comms teams now risk stamping their press releases with machine-readable AI markers under EU transparency rules, misrepresenting authorship to regulators and readers.
Ask SkimNews
