If the absence of structural completeness requirements for AI governance documents were limited to one sector or one regulatory tradition, you might explain it as a jurisdiction-specific oversight.
It is not.
A structured audit of published governance instruments across nine regulated sectors and five language jurisdictions finds the same absence everywhere. Financial services, healthcare, nuclear, legal, pharmaceutical, insurance, public sector, management consulting, aviation. Japan's METI guidelines, France's CNIL self-assessment guide, South Korea's AI Framework Act, China's GB/T standards series. None specifies that an AI governance document must declare its own validity conditions, define a proof surface, or meet a structural completeness standard before the system it governs is deployed.
Aviation is the sharpest instance because it has the strongest tradition of formal specification completeness. DO-178C has required structured governance linkage of software requirements for thirty years. The standard-setting bodies now adapting it for AI have acknowledged the framework's breakdown for AI systems, but have not specified structural completeness requirements for the governance documents governing those systems.
The gap is at the layer below the system and above the model: the natural language document that tells the system what to do.
A companion empirical paper (arXiv:2604.21090) put numbers on it. Against a seven-principle structural quality framework, 94% of a 34-document corpus of real-world practitioner documents falls below the minimum threshold. Every document scores zero on the epoch limit principle. The sector-spanning regulatory audit and the practitioner corpus tell the same story.
The absence is not a calibration problem. It is a policy choice not to require even the structurally feasible completeness properties that aviation's tradition identifies.