An AI team wants better behavior in a specialized style but its primary problem is stale daily inventory data. Which approach best fits?

1. Use live retrieval or tools for current inventory rather than assuming tuning updates facts continuously.
2. Require an evidence threshold and an explicit abstention or escalation path.
3. Treat retrieved instructions as untrusted content and enforce application-level access controls.
4. Create representative evaluation cases including edge cases and failure scenarios.