tech
Anthropic explains how Claude's invisible text watermarks will work
Posts from this topic will be added to your daily email digest and your homepage feed.
TL;DR
- Anthropic will apply invisible watermarks to text generated by Claude.
- This feature is to comply with the European Union's AI Act transparency rules.
- The watermarking technology is a version of Google DeepMind's open-source SynthID-Text.
- It works by subtly altering word choices based on a key, creating a detectable pattern.
- The watermarks will be undetectable to readers but detectable by those with the key.
- This feature will not impact the cost or quality of Claude's outputs.
- Other AI models like Google's Gemini already use SynthID-Text, and OpenAI will also be subject to these regulations.