Claude to start watermarking AI-generated text – but will it make quality worse?
Anthropic says it will change way chatbot makes small, random choices, to comply with EU regulation The world is familiar by now with the usual tropes of machine-generated text: overuse of the word “delve”, an excess of em dashes, and the chirpy, relentless construction of “it’s not X but Y”. But is it about to get even worse? Continue reading...
Open original article- Ledger record
- #20
- First seen
- 17 Aug 2026, 21:26 UTC
- Last checked
- 17 Aug 2026, 21:26 UTC
- Version history
- 1 version
- Current fingerprint
0d25304ff054...7ba11c
Publisher layer
Recorded article metadata
- Publisher
- The Guardian
- Country
- England / United Kingdom
- Source published
- 17 Aug 2026, 16:52 UTC
- Source updated
- No update timestamp supplied
- Byline
- Aisha Down
- Section
- Anthropic
Anthropic says it will change way chatbot makes small, random choices, to comply with EU regulation The world is familiar by now with the usual tropes of machine-generated text: overuse of the word “delve”, an excess of em dashes, and the chirpy, relentless construction of “it’s not X but Y”. But is it about to get even worse? Continue reading...
Change layer
Observed version timeline
-
1
Claude to start watermarking AI-generated text – but will it make quality worse?
Anthropic says it will change way chatbot makes small, random choices, to comply with EU regulation The world is familiar by now with the usual tropes of machine-generated text: overuse of the word “delve”, an excess of em dashes, and the chirpy, relentless construction of “it’s not X but Y”. But is it about to get even worse? Continue reading...
Changed fieldstitlebylinedescriptionTextcontentTextsourcePublishedAtsourceUpdatedAtsectionimageUrltags- Version SHA-256
0d25304ff054...7ba11c- Capture SHA-256
2bce3d4df01c...722802
Interpretation layer
Labels with declared origins
Publisher categories are copied from source metadata. Machine signals report observable wording or format and do not determine truth, intent, ethics, or wrongdoing.
- machine content-type Reported news observable-rules-v1 / 70% rule confidence
- publisher category AI (artificial intelligence) rss-category
- publisher category Anthropic rss-category
- publisher category Computing rss-category
- publisher category Internet rss-category
- publisher category Regulators rss-category
- publisher category Technology rss-category
Public response layer
Reception snapshots
Comment counts and samples remain absent until a lawful, source-specific capture method is reviewed. Absence is recorded instead of estimated.
Evidence layer
Capture provenance
- Capture ID
- #1
- Observed
- 17 Aug 2026, 21:26 UTC
- HTTP result
- 200 / parsed
- Raw payload
- 460,621 bytes, private gzip evidence
- Collector
- news-ledger/0.1
- Public boundary
- metadata and excerpt