Claude's invisible watermarks aim to prove AI processing under EU AI Act
TL;DR Summary
Anthropic will embed invisible watermarks in Claude outputs to comply with the EU AI Act, applying to new models globally and to processed content where supported, with a forthcoming detection API; but the marks can be bypassed, may mislabel human edits as AI-processed, and their effectiveness remains uncertain, with penalties up to €15 million or 3% of global revenue for noncompliance.
- Claude's new Scarlet Letter watermark is invisible — for now arstechnica.com
- Claude Users Can’t Opt Out Of New Watermarks—Here’s What We Know Forbes
- Anthropic's text watermarks signal new front in AI detection Axios
- Some Claude users are mad that Anthropic's new watermarks will catch them using it at their jobs, classes techcrunch.com
- Claude is watermarking every response — here's what that means if you use AI for writing Tom's Guide
Reading Insights
Total Reads
0
Unique Readers
5
Time Saved
8 min
vs 9 min read
Condensed
96%
1,645 → 61 words
Want the full story? Read the original article
Read on arstechnica.com