Summary
Anthropic is developing an invisible watermarking technology for AI-generated text, which embeds subtle statistical patterns to allow computers to detect content created by models like Claude. This technology aims to identify AI-generated content online and is compared with Google's SynthID and C2PA provenance.