Ecaveo Responsible AI Series
Commensurate Confidence
artificial intelligence, assurance and airworthiness in aerospace and defence
Fractional Head of AI, Ecaveo
Abstract
The customer’s policy does not bind you, the certification standard does not exist yet, and the most instrumented fleet in history is getting less available. Everything else follows from those three. JSP 936 binds the Ministry of Defence and does not bind its suppliers. The obligation reaches industry only where a contracting team writes it into a contract, which means a company waiting for the clause will be assembling evidence retrospectively about work it has already done. In civil aerospace the position is starker still, because no published certifiable process standard for machine learning exists, the joint EUROCAE and SAE standard remains in draft with a target of end 2026, and its first issue will be capped at design assurance level C and will exclude adaptive learning. This paper argues that the assurance case is the product. A company that builds the evidence chain as it engineers, from need through configuration and test into service, can sell into both markets from one controlled record. A company that builds demonstrators will keep building them.
Read the paper
This is the reading edition, complete and unabridged. The designed edition, with the full illustrated layout, is available from the original publication page.
Cite this paper
BibTeX
@techreport{forrest2026commensurate,
author = {Forrest, Paul},
title = {Commensurate Confidence: artificial intelligence, assurance and airworthiness in aerospace and defence},
institution = {Ecaveo Services Ltd},
series = {Ecaveo Responsible AI Series},
year = {2026},
month = {07},
pages = {39},
url = {https://papers.ecaveo.com/commensurate-confidence-ai-in-aerospace-and-defence/}
}