The WHO Principle and the Amber Formula: A Two-Step Method for Verifying Claims in AI-Assisted Work

Large language models produce fluent, confident, well-formatted text regardless of whether the underlying claim is true. This paper describes two simple, sequential checks — the WHO Principle and the Amber Formula — developed and tested during the operation of a real company, for catching hallucination before it reaches a document, contract, or public claim. Three real cases are documented. The method is deliberately minimal: two short questions and one three-part filter, applied at the point where a claim is about to be acted on.