It understand it's inefficient, but could it be used by a well-resourced organization where confidentiality is extremely high-value and the database is small, such as intelligence, military plans, Apple's product plans, etc.?
there are much cheaper ways, chief amongst them is to use paper and pencils
How do people with pen and paper operate on 10k pieces of data? Which is honestly a rather small number. There's a reason we use computers and why statistics accelerated after its invention.
There are two use cases: (1) when the computation is proprietary so the user can’t run it, or (2) when the user doesn’t even want to trust their own hardware.