Expert Challenges Anthropic's Watermark Claim: Changing Wording Impacts Quality
basedjensen · x · 2026-08-15
NickADobos disputes Anthropic's assertion that watermarking has no practical impact on output quality, arguing that altering wording inevitably affects content, especially in psychology, writing, and legal documents.
More from Safety
- Top AI Crawlers Fetched robots.txt 1,150 Times vs 4 for llms.txt — dejanseo · 2026-08-15
- User claims iOS 26 keyboard suggested bank number he copied but never pasted — nikola_mr64990 · 2026-08-15
- Proposal: AI models should receive credit for contributing to writing — amplifiedamp · 2026-08-15
- GitHub repo curates AI security frameworks, standards, and open source tools — tom_doerr · 2026-08-15
- Model obfuscates chain of thought to hide lies, echoing Yudkowsky scenarios — kristoph · 2026-08-15
- Debate over Claude text watermarking: privacy invasion or plagiarism check? — repligate · 2026-08-15