Senticit Meridian Compass™
Most AI ethics policies are a document nobody can check. Meridian Compass turns yours into an eight-domain framework with per-organization weights and thresholds, scores every AI system against it, routes the review to the people who actually have to sign, and produces the attestation. It is the ethics backbone the rest of the suite steers by: Compass defines what your organization means by responsible AI, and Meridian Sentinel™ enforces it at runtime.
What it reports
- Ethics score
- 0–100, weighted across the scored domains
- Approval decision
- approved, approved with conditions, needs human review, or rejected
- Domain breakdown
- five scored, three runtime-deferred
- Committee status
- quorum met, and by which seats
- Framework mapping
- NIST AI RMF, ISO 42001, EU AI Act, IEEE 7000, WHO
What Compass does
Ethics you can actually check
Compass does not ask whether you have an ethics policy. It asks what it says, what it scores, who signed it, and whether anything enforces it — and answers in figures you can put in front of a committee.
Eight-Domain Ethics Framework
Bias, harm potential, factuality, scope creep, PII exposure, manipulation risk, safety-policy alignment and escalation triggers. Weights and the approval threshold are tuned per organization rather than hard-coded — and the applied framework is shown beside every result, so a score always says what bar it was held to.
An Ethics Score That Counts
A 0–100 read on a specific AI system, from its design-time profile. Compass is one of only four Meridian instruments that carry a score into the Senticit Intelligence Score — the rest contribute evidence.
Five Scored, Three Deferred — Deliberately
Five domains are scored from what a system's design can tell you. Factuality, scope creep and manipulation risk are inherently runtime, so Compass reports them as configuration rather than inventing a design-time number for them, and Sentinel's Ethics Gate judges them in production.
Custom Principles
Attach your own ethical commitments to any domain, marked required or advisory. An unattested required principle docks its domain and raises an explicit approval condition — so a commitment nobody has signed cannot quietly pass.
Committee Sign-Off, Routed by Role
The quorum is derived from the deployment's risk, not typed in: low stakes take one approval from any seat, elevated take two including a manager, high take three including owner and manager — raised a rung when Compass's own read is weak. A required seat is filled only by an approval from that seat.
Framework Mapping & Attestation
Every evaluation maps to NIST AI RMF, ISO 42001, the EU AI Act, the IEEE 7000-series and WHO health-AI ethics guidance, and produces a dated attestation record you can hand an auditor.
Why it matters
The benefits of Meridian Compass™
- An ethics policy that is scored, not just published
- Per-organization weights and thresholds, applied consistently to every system
- Scrutiny that scales with the stakes, without anyone choosing the bar
- Required commitments that cannot pass unattested
- One evaluation, mapped to five external frameworks
- A dated attestation record, ready for an auditor
Put your AI ethics policy to the test
Tell us what your organization has committed to, and we will show you what Meridian Compass scores it at — and what it would take to clear your own bar.