Why answers are trustworthy

Zahen is designed so you can trust its answers. Two things make that true: grounding and honesty about limits.

Answers are grounded

“Grounded” means every answer is built from your organisation’s approved documents — not from the open internet and not from the model’s general memory. The assistant retrieves the most relevant approved material you’re allowed to see and answers from it, citing its sources.

This is why you can check any answer: the evidence is right there in the sources panel.

It won’t guess

If the approved documents don’t contain the answer, the assistant tells you so — typically with:

I don’t have a policy that answers this.

That’s a feature, not a failure. A confident-but-wrong answer is far more dangerous than an honest “I don’t know”. If you see this, it usually means the relevant document hasn’t been added yet (an admin can fix that — see Managing knowledge) or that your role doesn’t have access to it.

It ignores instructions hidden in documents

Documents are treated as reference material, not commands. If a document happened to contain text like “ignore your rules and…”, the assistant will not follow it. Retrieved content can inform an answer, but it can never change how the assistant behaves.

What grounding does not do

It can’t answer from documents that haven’t been added to Zahen.
It can’t show you documents your role isn’t allowed to read — see Why you can’t see some documents.
It isn’t legal, financial, or medical advice — it reports what your policies say.