Blog

AI, data and trust

The two questions every buyer asks before approving an AI build are whether their data trains someone else’s model and how they would know if the system started making things up. Both have precise answers, in vendor contracts and in published evaluation methods, and both are usually answered with reassurance instead.

Is your RAG system lying
Four different failures look identical to a user. Each has a different fix, and each is separately measurable.
5 October 2026
Does your data train the model
Every major provider commits not to train on business API data. The trap is a free tier on the same endpoint.
4 October 2026

AI, data and trust

Is your RAG system lying

Does your data train the model