Skip to content
Indem Insure

Indem research / Open framework / v1.0

Physical AI InsuranceReadinessIndex.

A 100-point framework for assessing whether an autonomous fleet or robot can produce clear, current, source-linked evidence for insurance review.

Download the scorecard The methodology

Evidence readiness. Not a risk rating.

100possible points
5evidence dimensions
20scored indicators
1reusable risk record

01 / What it measures

Evidence readiness,
not a promise.

Physical AI changes after deployment. Hardware is serviced, software is released, operating domains expand, and incidents create new information. The Index measures whether that changing history can be followed by another party without relying on a founder's memory.

Evidence readiness, not a promise of safety or coverage. The Index does not estimate expected loss, certify a system, determine legal compliance, or guarantee a quote. It organizes evidence so operators, brokers, underwriters, and risk partners can ask better questions.

0Absent

No usable evidence

1Informal

Known by people, not reliably recorded

2Partial

Some records, with material gaps

3Documented

Repeatable records with ownership

4Source-linked

Records trace back to primary evidence

5Continuous

Evidence stays current as the system changes

02 / The scorecard

Five dimensions.
One living record.

01

System identity and configuration

Can a reviewer identify exactly what system was operating?

4 indicators / 20 possible points
  1. Asset inventory with stable vehicle, robot, hardware, and unit identifiers
  2. Safety-relevant software, model, firmware, and map versions by unit
  3. Sensor, compute, communications, and fallback configuration
  4. Dated deployment, configuration-change, approval, and rollback history
02

Operating domain and exposure

Can the business show where, when, and how the system works?

4 indicators / 20 possible points
  1. Defined operating domain, including geography, road, facility, task, and weather limits
  2. Mileage, operating hours, trips, tasks, or cycles measured by unit and period
  3. Exposure segmented by route, customer, cargo, passenger, task, or use case
  4. Recorded exceptions, out-of-domain events, overrides, and abnormal conditions
03

Human oversight and controls

Can a reviewer see who remains accountable when autonomy acts?

4 indicators / 20 possible points
  1. Named owners for deployment, operations, safety, maintenance, and incident response
  2. Remote assistance, intervention, takeover, and override events retained with context
  3. Operator, technician, and safety personnel qualifications and training records
  4. Escalation, shutdown, minimum-risk, and emergency-response procedures
04

Maintenance and change management

Can the business prove that the system is cared for as it changes?

4 indicators / 20 possible points
  1. Preventive maintenance requirements and completion history by unit
  2. Inspection, sensor calibration, repair, and return-to-service records
  3. Defect, recall, degradation, and remediation history
  4. Release validation, change approval, monitoring, and rollback controls
05

Incidents and outcomes

Can events be reconstructed and used to reduce recurrence?

4 indicators / 20 possible points
  1. Consistent taxonomy for incidents, collisions, near misses, and safety interventions
  2. Event timelines linked to telemetry, video, logs, statements, and preserved source files
  3. Injury, property damage, third-party, regulatory, claim, and loss outcomes
  4. Root-cause findings, corrective actions, owners, deadlines, and recurrence monitoring

03 / Reading the result

A maturity signal.
Not a decision.

0-39

Discovery

Critical evidence is missing or informal.

40-59

Structured

Core records exist, but important gaps remain.

60-79

Reviewable

A reviewer can follow the risk with targeted follow-up.

80-100

Continuously evidenced

The evidence package is current, traceable, and reusable.

Score every indicator from 0 to 5, then add the 20 scores. Review the lowest-scoring indicators before focusing on the total. A high score means evidence is easier to inspect; it does not mean the underlying risk is low.

04 / Public reference points

Real-world evidence.
Public foundations.

NISTAI Risk Management Framework

Continuous governance, mapping, measurement, and management across the AI lifecycle.

NHTSAStanding General Order on ADS and ADAS crash reporting

A current federal example of structured incident reporting for identified manufacturers and operators.

FMCSAAutomated CMV Evaluation program

Operational research on automated and cooperative driving systems for commercial motor vehicles.

The Index is an independent Indem framework. It is not issued or endorsed by NIST, NHTSA, FMCSA, any insurer, or any regulator.

From the framework to your operations

Make it reviewable.

Start a Physical AI pilot Return to Physical AI

Suggested citation

Indem. 2026. Physical AI Insurance Readiness Index, Version 1.0.https://indeminsure.com/physical-ai/readiness-index/