Heidi is built so she does not have to. Answers come from sourced facts, weak facts stay out of answers, and when she does not know, she says so instead of guessing.
Hallucination is what happens when a model must answer without knowledge. Heidi's answers are grounded in the Cortex: facts with sources and confidence scores. A fact that is weak or contested stays out of the answer.
And the honest fallback is built in: if she does not know, she says so. She never makes things up is a product rule, enforced by grounding, not a hope.