Anthropic opens Claude text-watermark detection to outside verifiers

Anthropic launched an API letting regulators, media outlets, fact-checkers, and researchers check whether text carries Claude's invisible digital watermark. The move responds to the European Union AI Act's requirement that AI-generated text carry such markers. Critics argue the technology could degrade text quality and create problems where contracts prohibit AI use.

Why it matters: This is one of the first concrete moves by a major lab to operationalize AI-content provenance for outside verifiers rather than just internal use, testing whether watermarking actually works at scale under real regulatory pressure. It sets a template other labs may need to follow as the EU AI Act's transparency requirements take effect elsewhere.