Because large language models are the classifiers creating the risk.They generate probabilistic interpretations.They do not enforce deterministic release decisions.Preflight Clearance exists upstream of models — as a governance boundary.Models produce uncertainty.Preflight Clearance controls commitment.LLMs cannot govern themselves.
Each evaluation produces a written assessment document. This document includes:
The assessment is delivered within the agreed timeline, typically three to five business days depending on document length and complexity.
Formal evaluation begins before publication.