An AI Watermark Won't Tell You What to Trust
Anthropic now marks everything Claude writes. But a positive match tells you a machine was involved, not what it did — and the harder question was never authorship. It's when to believe an answer, from a model or from anyone else.