DEVOURED
Claude's Text Output Now Carries an Invisible Watermark
Anthropic is now embedding invisible SynthID-Text watermarks into Claude's outputs to comply with European Union AI regulations.
What: Anthropic updated its Claude models to include an invisible watermark that nudges word choices during generation to aid in content detection without affecting text quality or adding hidden tokens.
Why it matters: This move marks the beginning of standardized provenance protocols for AI-generated content required by emerging global regulations like the EU AI Act.
Decoder
- SynthID-Text: A digital watermarking technology that modifies the probability distribution of token selection to create a detectable pattern without significantly altering the generated text.
Original article
Anthropic's latest Claude models embed a SynthID-Text-based watermark in all text output to comply with the EU AI Act. It works by nudging choices among near-equivalent words, adds no hidden characters or tokens, and carries no information about the user.
Explore all 55 articles in today's digest
Fresh dev news, AI-digested daily. Never miss a trend at devoured.fyi