CAS Compass

CAS Compass: audit AI behavior and inspect the evidence.

An early-access surface for scoped model-behavior audits. Findings are tied to traceable evidence, stated limitations, and explicit human-review boundaries.

Customer-production gate: OPEN

Research evidence does not automatically become ACCEPT. Items that require judgment remain FLAG_FOR_HUMAN_REVIEW.