Redditor argues local-model translation could strip Claude's text watermark
Some_Kitchen7030 · reddit · 2026-10-09
A Reddit user walks through a theoretical bypass of Anthropic's probabilistic text watermarking: write an essay in Thai with Claude, then translate it to English with a local open-source model. Their argument: a local model won't adhere to the "green set" word subset Claude uses for tracking, structural watermarks would also be rewritten, and local models aren't bound by vendor policies — making the output very hard to detect.
More from Safety
- Super-Worms Warning: AI-Powered Self-Propagating Malware and a Fragmenting Internet — JensHonack · 2026-10-23
- Claude Code Pushes Urgent Security Update Across All Installations — umeshonai · 2026-10-23
- Sanders calls to pause autonomous AI agents in bank accounts; critics push back — AIandDesign · 2026-10-23
- TechPolicy essay: system prompts are weak evidence of AI safety, outputs must be evaluated — mmitchell_ai · 2026-10-23
- Anthropic's deputy security chief warns cyber threats will get 'a lot worse before they get better' — Polymarket · 2026-10-23
- AI cyberattacks could paradoxically accelerate AI development in an offense-defense arms race — kimmonismus · 2026-10-23