Using AI To Write? You Could Be Caught! Anthropic's Invisible Watermark Explained

Think AI-written content can't be detected? Anthropic is working on invisible watermarking for AI-generated text that could allow computers to identify content created with Claude.

The watermark isn't visible to the reader. Instead, Anthropic says it can be embedded through subtle statistical patterns in the way the AI generates text. But how reliable is it, and can it survive copy-pasting or editing?

We explain how AI text watermarking works, how it compares with Google's SynthID, what C2PA provenance means and why this technology could become important for identifying AI-generated content online.