AIHacker Newsabout 2 hours ago
Anthropic Watermarks All Claude Text Output
1 min read
Anthropic will watermark all Claude-generated text worldwide using semantic steganography, altering word choices at inference time to embed detectable fingerprints. Only Anthropic can detect Claude's watermarks, and detection confidence scales with text length. The technique changes word selection, potentially affecting precision of generated writing.
Level
Hype check
Tap to vote and see what everyone thinks.
In this storyAnthropic
