Due diligence on AI vendors: ten dimensions, and what the Builder.ai collapse teaches

In short: AI vendor due diligence means scoring each supplier on ten dimensions (capability on your task, security, data use, compliance, transparency, reliability, controllability, cost, viability and exit) using only evidence you hold, not sales claims. The collapse of Builder.ai in 2025 showed why viability, transparency and exit belong on the list.^1,2,3

Buying AI is buying a dependency: on a model you did not train, a supplier you did not audit and a roadmap you do not control. ISO/IEC 42001’s control objective on third-party and customer relationships and the NIST AI RMF’s supply-chain guidance ask for the same thing in formal language: know what you are buying, from whom, and on what evidence.^1,2