The Rule · v0.1 · September 2026
One role per model. We build it, or we assure it. Never both.
For any given model, The Key 4 AI takes one of two roles. Either we build it, train it and stand behind it, or we independently evaluate, validate and assure it. We do not do both for the same model, for the same client, at any time.
The reason is simple. A verifier is only worth something if it does not care whose model it is grading. The moment the grader has a stake in the result, the grade is worth nothing, however good the engineering.
In practice: before any engagement starts, the role is written down. If we built a model, its independent assurance is done by someone else, and we will help you find them. If we assure a model, we will not later be paid to improve it.
This rule binds the company, not just a project. It is versioned here, and any change will appear here first, dated.
Signed for The Key 4 AI, Kochi. v0.1.
Honest enough to test.
We claim nothing you cannot check for yourself, and we never grade our own work.One address