AI Text Watermarking Is Neither New Nor Permanent; Google Shipped It Over a Year Ago
EXM7777 · x · 2026-08-13
In response to Claude's upcoming invisible watermarking, a user notes that AI text watermarking is not new; Google shipped it over a year ago with Gemini 2.5. The watermark can be removed via simple tricks from GitHub, such as manipulating Unicode characters and patterns. The author argues the news is overblown, as readers only care about content quality, not whether Claude helped write it.
More from AGI Musings
- Researcher Slams LLMs for Causal Inference: Fabricating Data Like Fan Fiction — RexDouglass · 2026-08-13
- AI Reshapes Academia: Mathematics Becomes Capital Intensive — analisereal · 2026-08-13
- AI Shifting Science: Mathematics May Become Capital Intensive — analisereal · 2026-08-13
- Viral AI Joke: Once Machines Solve Physics, Humans Are Free for Condensed Matter Physics — fkasummer · 2026-08-13
- Ex-OpenAI/DeepMind Safety Lead: We Have 3 Years to Solve Superintelligence Alignment — 141_1337 · 2026-08-13
- Big Tech's AI Growth Relies on OpenAI, Masking Unreal Revenue Loops — GaryMarcus · 2026-08-13