Aviation certification reveals 37% of AI governance docs lack structural quality

A new academic paper from Christo Zietsman draws on 50 years of aviation software certification to critique the structural weaknesses in current AI governance. The study highlights that aviation standards like DO-178C and DO-330 have required three structural properties since 1992: structured governance linkage between specifications and evidence, context-bounded validity that triggers revalidation when operational context changes, and an objective evidence architecture defining what proof means. These properties are enforced by FAA and EASA certification, yet no existing AI governance framework—system prompts, governance policies, or task envelopes—requires any of them.

The paper maps these aviation requirements onto three structural findings for AI: epoch limits on governance document validity, proof surfaces as revalidation feedback, and the absence of structural completeness in AI governance instruments. An empirical companion study found that 37% of AI governance documents fall below the structural quality threshold. To address this, the PromptQ framework operationalises the three aviation requirements at the governance document layer. The research concludes that while aviation's system-level requirements break down for non-deterministic AI, the document-level structural properties remain fully transferable, offering a clear path to improve AI governance rigor.