The design assumption behind every major AI disclosure proposal is that the bad actor is an outsider — a troll farm, a foreign state, a scammer — and that institutions represent the verification layer. That assumption collapsed when the institutions became the source. The combination of real and synthetic footage as official communication is not a bug in a foreign influence operation; it is a feature of domestic political messaging that existing law does not touch. South Africa's experience, where hallucinated citations in a ministerial AI policy required the document's…