Burkov slams watermarking in paid LLM outputs: 'I paid for this text'
burkov · x · 2026-10-08
Andriy Burkov, author of The Hundred-Page Machine Learning Book, complains that frontier closed-weight LLMs inject statistics-based watermarks into the text they generate for paying users. He argues that paid-for text belongs to the user and shouldn't be marked in any way, and is convinced that watermarked text, once detected on a webpage, will hurt its search ranking or even get an entire site shadowbanned by search engines.
Related event: ML Author Protests LLM Watermarks, Releases Dehydrate Removal Tool(2 posts)→
More from Models
- Datalab's Chandra OCR 2 Model Now on Hugging Face — VikParuchuri · 2026-10-08
- Codex and ChatGPT Work hit 40M active users as OpenAI hands out banked resets — kimmonismus · 2026-10-08
- New collection of open-source decision and embedding models hits near-SOTA — antoine_chaffin · 2026-10-08
- Haiku 5.5 hits 1578 Elo on AA-Briefcase, within confidence range of GPT-6 Astra — ArtificialAnlys · 2026-10-08
- Claude Haiku 5.5 burns ~162k tokens per task, 3x GPT-6 Luna for similar scores — ArtificialAnlys · 2026-10-08
- OpenAI's Dots: always-on agents inside ChatGPT, powered by GPT Astra — thursdai_pod · 2026-10-08