# Faultprint Operating Diagnostic: USAA SafePilot

Run ID: FP-OPERATING-0014
Mode: Operating Mode
Institution: USAA
Corpus: Public SafePilot and DriveSafe support pages
Diagnostic track: Source-of-Truth Collapse
Report posture: Findings only. No recommendations.

## Executive Read

The reviewed corpus supports monitored exposure around app-based driving score, renewal discount calculation, phone-handling classification, and policy-level aggregation across rated drivers. The most distinctive condition is that USAA states phone-handling infractions may be recorded even if someone else uses the phone. The report does not classify failure because the corpus lacks dispute outcomes, correction records, policyholder samples, validation reports, or complaint evidence.

### Finding USAA-F01: SafePilot discounts are based on driving behavior at renewal.

Finding classification: Exposure.

Evidence Anchor

- Source: USAA SafePilot driving score discounts page.
- Corpus locator: `Driving discounts calculation`.
- Evidence excerpt or bounded evidence note: USAA states SafePilot can earn up to a 30 percent policy discount at renewal and that the score is locked before renewal.
- Claim type: Source Fact.
- Analytic bridge: Renewal discount calculation makes the score consequential for policy cost.
- Confidence: High.
- Falsification test: Policy records showing SafePilot scores do not affect renewal discounts would downgrade exposure.
- Boundary: This does not prove adverse pricing or harm.

### Finding USAA-F02: Earned driving discount is based on all rated drivers on the policy.

Finding classification: Exposure.

Evidence Anchor

- Source: USAA getting-started SafePilot page.
- Corpus locator: `SafePilot and SafePilot Miles note`.
- Evidence excerpt or bounded evidence note: USAA states the earned driving discount is offered at renewal and is based on driving behavior of all rated drivers on the policy.
- Claim type: Source Fact.
- Analytic bridge: Aggregating multiple drivers can create dependency between one person's behavior and another person's policy outcome.
- Confidence: High.
- Falsification test: Rating records showing driver behavior is isolated and cannot affect other rated drivers would downgrade exposure.
- Boundary: This does not prove unfair aggregation.

### Finding USAA-F03: Phone-handling infractions may be recorded even if someone else uses the phone.

Finding classification: Exposure.

Evidence Anchor

- Source: USAA DriveSafe app FAQ.
- Corpus locator: `Phone handling FAQ`.
- Evidence excerpt or bounded evidence note: USAA states the app will record phone-handling infractions even if it is not the policyholder using the phone.
- Claim type: Source Fact.
- Analytic bridge: This creates an explicit attribution-risk condition where the measured event can diverge from the responsible driver behavior.
- Confidence: High.
- Falsification test: Correction logs showing such events are reliably removed before scoring would downgrade exposure.
- Boundary: This does not prove the classification causes harm or is uncorrectable.

### Finding USAA-F04: SafePilot Miles combines usage and behavior discount logic.

Finding classification: Exposure.

Evidence Anchor

- Source: USAA SafePilot Miles page.
- Corpus locator: `Usage Based Insurance description`.
- Evidence excerpt or bounded evidence note: USAA states customers can receive a discount for driving less and an additional renewal discount for safe driving.
- Claim type: Source Fact.
- Analytic bridge: Combining usage and behavior creates a dual-source pricing surface.
- Confidence: High.
- Falsification test: Product rules showing usage and behavior are never combined in policy pricing would downgrade exposure.
- Boundary: This does not prove the combination is unfair or inaccurate.

### Finding USAA-F05: No public correction breakdown is evidenced.

Finding classification: Null finding.

Evidence Anchor

- Source: Reviewed USAA SafePilot public corpus.
- Corpus locator: `Corpus-level review`.
- Evidence excerpt or bounded evidence note: The reviewed corpus does not include phone-handling dispute outcomes, driver-attribution correction samples, validation audits, or complaint-to-correction records.
- Claim type: Analytic Inference from corpus absence.
- Analytic bridge: Without correction failure evidence, this run cannot classify collapse.
- Confidence: Medium.
- Falsification test: Records showing recurring unresolved phone-handling or driver-attribution disputes would eliminate this null finding.
- Boundary: Absence from public materials does not prove absence inside USAA.

## Scope Calibration

The corpus supports exposure around policy-level behavior scoring, renewal discounts, phone-handling attribution, and usage-plus-behavior logic. It does not support claims of failure, harm, or ineffective correction.

This report makes no recommendations.