Methodology · public document

APIR Insurance Score™

The 0–100 rating that bridges AI compliance posture and risk pricing, the underwriting data layer for E&O, professional liability, and cyber policies covering AI deployments. APIR produces the rating and the evidence; licensed carriers and brokers price and bind any policy, not APIR.

Build your first agent free

Eight sub-scores

Behavioral Reliability

Weight 20%

Action verification rate, rule compliance, high-risk action share

Incident History

Weight 20%

Critical + open incidents, resolution record

Compliance Coverage

Weight 15%

Framework compliance status (EU AI Act · NIST · ISO 42001)

Drift Stability

Weight 15%

Behavioral drift, fingerprint divergence, audit anomalies

Containment Readiness

Weight 10%

Ghost Audit coverage, active monitoring, audit depth

Data Governance

Weight 8%

Trust score contribution, platform registration, compliance

Transparency

Weight 7%

Traceability: platform, model identity, fingerprint baseline

Human Oversight

Weight 5%

Decision authority model, human review rate, monitoring

Scoring formula

Each sub-score is 0–100. The Insurance Score is the weighted blend, rounded to one decimal:

insurance_score = round₁( Σᵢ (sub_scoreᵢ × weightᵢ) )    where Σᵢ weightᵢ = 1.00

Rating-agency grade bands map the score to an evidence band a broker can price against:

AAA

≥ 90

Strongest evidence coverage

≥ 85

Very strong

≥ 80

Strong

BBB

≥ 75

Good

≥ 70

Adequate

≥ 65

Minimum evidenced threshold

CCC

≥ 55

Weak evidence coverage

≥ 45

Very weak

≥ 35

Severely limited

≥ 0

Not evidenced

Insurability tiers & certification

The score also maps to an insurability tier, the vocabulary brokers see on every APIR surface:

Standard≥ 80Insurable: verify GL AI exclusions (CG 40 47/48) with your broker

Substandard≥ 65Insurable with elevated premiums and coverage restrictions

High Risk≥ 50Significant restrictions: specialized AI liability cover required

Restricted≥ 35Generally excluded: remediation required before cover

Uninsurable≥ 0Below minimum threshold: not currently insurable

Certification threshold: 65. Agents at grade B or above are eligible for an APIR Evidence Certificate, issued only by a human-triggered call, valid 365 days, publicly verifiable. It attests what the record contained on the day it was issued. It is not a statement about insurance.

Note: the APIR Insurance Score is 0–100. The FICO-style 300–850 scale belongs to the separate APIR Agent Credit Bureau product.

Worked example

An agent with:

behavioral_reliability:  82  (weight .20)
incident_history:        75  (weight .20)
compliance_coverage:     90  (weight .15)
drift_stability:         85  (weight .15)
containment_readiness:  100  (weight .10)
data_governance:         80  (weight .08)
transparency:            70  (weight .07)
human_oversight:         88  (weight .05)

Produces:

insurance_score = 82×.20 + 75×.20 + 90×.15 + 85×.15 +
                  100×.10 + 80×.08 + 70×.07 + 88×.05
                = 16.4 + 15.0 + 13.5 + 12.75 + 10.0 + 6.4 + 4.9 + 4.4
                = 83.4  →  Grade A  ·  insurability: standard

The worked example produces 83.4, which is Grade A. What that number means for any policy is a question for a licensed broker. APIR does not price, place or assess cover, and has no underwriting partner.

Reproducibility

Every agent_insurance_scores row carries the methodology_version that produced it, and every rescore appends to a tamper-evident score history. A broker reading a score you share knows exactly which version of this rubric was applied; historical scores stay comparable even after methodology updates. One engine computes every score, there is no second code path.

Citing this document

APIR Intelligence. (2026). APIR Insurance Score Methodology v1.1.0.
https://apir.ai/insurance/methodology

← apir.ai Trust Genome methodology