RATSe Calculator updated July 2026

Resilient AI Trust Score — environmental

RATSe Governability

LEVEL 1 · ASSESSMENT  A deterministic, non-compensatory maturity rubric for AI governance readiness — design-time and periodic. This produces an assessment score, not a runtime control signal; RATSe Runtime (Level 2) is the separate live-control layer, and this score must never be wired into a runtime gate.

Master rubric — 6 pillars

Overall governance level · weakest link

—

Set scores and evidence at left. The readout updates live.

Performance Index

— /100

benchmark only — not the verdict

Evidence · weakest

——

lowest evidence grade present

Pillar vectors — live status

Pillar breakdown

Pillar profile

Export

How it scores

Score 0–3 per criterion: 0 absent · 1 declared (theatre) · 2 defined/partial · 3 enforced & verifiable. Evidence A–D: A third-party · B system-logged · C self-attested · D none.

Pillar score = 100 × Σ(score×risk-weight) ÷ Σ(3×risk-weight). Weights: Low 1 · Med 2 · High 3 · Critical 5. Levels: <40 Foundational · 40–69 Evolving · ≥70 Advanced.

Three hard caps (non-compensatory):

Critical failure — any Critical criterion scored 0 forces its pillar to Foundational.

Evidence–competence — a score of 3 with evidence C or D counts as 2. No self-attested "Advanced."

Participation gate — if affected-party voice scores <2 or is weakly evidenced, the whole Ethical pillar is capped at Foundational.

Overall = the weakest pillar level, never an average. Evidence reported = the weakest grade present.

Self-assessment instrument. Output is Tier 1 / at most EAL C until independently validated. Do not present as an audit.

Message