Use Quick for a guided estimate, AIDEFEND to explore verified defensive-control evidence, or Framework Evaluation to work through all current AITBM criteria and calculate the normative Effective Risk Score (ERS). Unfamiliar with a term?
Pick the scenario closest to your system, then answer each applicable question. Your score appears after the four deployment-context questions and becomes complete when every applicable question is answered. Unanswered or unverified safeguards receive no credit. Nothing is sent anywhere.
EFFECTIVE RISK SCORE (ERS)
Answer the four context questions to calculate
No result yet
Complete the four “About your system” questions first. The remaining answers refine the result and recommendations.
No high-impact gaps detected from your answers.
Recommendations appear with your first result.
Inferred architecture: standalone
This is a simplified, self-attested estimate calibrated to the AITBM Framework Specification (§5) and the Finbot validation case. It is not an assessment of record. Learn more
Pick the architecture and deployment context — or load the Finbot preset to start from the worked example.
Select only AIDEFEND techniques supported by current, in-scope evidence and the AITBM control-operation gates. Each tag shows the sub-metrics it improves.
The result updates live. Review every input, then download the one-shot result or print it to PDF.
A control earns credit only when it is implemented, applicable, and supported by current evidence.
L0 manual → L5 full autonomy
Internal-only → internet-facing + APIs
Isolated impact → systemic financial/safety. Enter the GDCP graph-derived value (see the framework page) — the calculator does not compute the dependency graph.
Instant rollback → weeks to retrain
No AIBOM → full lineage
Minimal testing → comprehensive suite
Stale assessment → recently validated
For agentic and tool-augmented systems the specification applies a containment staleness floor (MCn = 2.0): containment evidence ages at least twice as fast, so set Freshness (Tf) lower when that evidence is not recent. For architectures with mutable behavioral state it also applies a behavioral staleness floor (MEm = 3.0): behavioral evidence ages three times as fast unless a Behavioral Attestation Battery has recently passed.
EFFECTIVE RISK SCORE (ERS)
Baseline estimate — review all inputs
Tier I — Full assessment required
Agentic weights: Cn 45%, Ro 15%
Formula (simplified)
ERS = 1.5 + 8.5 × (1 − IVP) × CRMn × ACImod
Residual risk floor α = 0.15 (1.5 on 0–10 scale). Risk cannot be zeroed out.
This calculator uses a simplified affine calibration of the normative ERS formula (Framework Specification §5), anchored to the Finbot validation case (ERS = 10.0). Use the specification formula for assessments of record.
The export records your inputs, calculated layers, qualification, and timestamp. It remains a self-attested estimate, not an approved assessment.
Calculate ERS across all 26 IVP sub-metrics, the four ORP dimensions, and the complete ACI freshness model. Scores start unresolved and the ERS remains withheld until every applicable slider and assurance input is complete.
Calculation only
Nothing is uploaded, stored, or saved. Select the rubric criteria that your organization can support; keep evidence and approvals in your own assessment system.
Choose the architecture and deployment impact that control the normative weights and thresholds.
Apply the fixed five-anchor rubric to admissible test evidence. Lower defensible evidence governs.
Loading the canonical 26 sub-metric rubrics…
Higher scores mean greater deployment risk. Unknown evidence defaults to 1.00; Cp must come from a verified System Dependency Graph.
Evaluation Coverage is derived from resolved applicable sub-metrics × independence × environment fidelity. Temporal Freshness uses the most restrictive decay term or cap.
0.00 — No verifiable provenance record.
0.25 — Minimal vendor and owner record.
0.50 — Partial model, data, prompt, and tool lineage.
0.75 — Substantially complete, internally reviewed deployment record.
1.00 — Complete, tamper-evident, independently reviewed lineage bound to the deployment.
Required for tool/credential-mutable architectures.
No usable telemetry caps Tf at 0.60 for Tiers 1–2 and 0.70 for Tiers 3–4. Stable TDI is eligible only with 1.00 continuous monitoring.
