Loss scenarios

Four losses a personal auto policy can't answer.

These are not edge cases. They are the ordinary shape of a claim once software is doing the driving — and in each one, the question of who pays turns on machine evidence rather than on what anyone remembers.

ADAS · Level 2–3
Signal: contested control

The handoff collision

  1. 01
    ScenarioA driver-assist system requests takeover approaching a work zone. The driver responds late, or responds to the wrong thing. The vehicle leaves its lane and strikes another car.
  2. 02
    Machine evidenceControl state second by second, the takeover request and its timing, driver-monitoring output, torque on the wheel, and whether the system was inside its declared operating domain when it asked.
  3. 03
    Where liability landsSplit, and contested. Personal auto responds for the driver's share; product liability responds if the handoff itself was unreasonable — too little warning, or a request the system should not have needed to make.
  4. 04
    What we rate onTakeover lead time, driver-monitoring quality, ODD boundary discipline, and observed intervention rates across the fleet.

Why it matters: this is the zone where a human is nominally responsible and functionally not — and the one carriers price most crudely.

Driverless fleet
Signal: no driver to name

At-fault, with nobody in the seat

  1. 01
    ScenarioA rider-only vehicle yields incorrectly at an unprotected left and is struck. There is no operator on board, and the other party names your company rather than a driver.
  2. 02
    Machine evidencePerception output, the planner's decision trace, the software version and map release in force, sensor health at the moment, and any remote-assistance session open at the time.
  3. 03
    Where liability landsOn the deployer, immediately — then redistributed across the stack. The autonomy vendor, the mapping provider, and the sensor manufacturer each sit behind a contractual indemnity that has to actually respond.
  4. 04
    What we rate onMiles by domain, incident and near-miss rates, remote-assistance load, and how cleanly the indemnity chain is drafted behind you.

Why it matters: a commercial claim against a technology company, priced as if it were a fender-bender, is a mispriced claim.

OTA release
Signal: fleet-wide, overnight

Behaviour regression after an update

  1. 01
    ScenarioA release ships on a Tuesday. By Friday, braking behaviour in low sun has changed just enough to produce a cluster of rear-end claims across several cities at once.
  2. 02
    Machine evidenceThe version manifest per vehicle, rollout timing by cohort, pre- and post-release behaviour on matched routes, regression test coverage, and time to detect and roll back.
  3. 03
    Where liability landsSquarely on the software provider — and on one event, not many. Every affected vehicle traces to a single defective release, which is the definition of a correlated loss.
  4. 04
    What we rate onStaged rollout discipline, canary cohort size, regression coverage, rollback time, and how much of the fleet ever runs one version at once.

Why it matters: this is the aggregation event traditional books are least ready for. An annual underwriting cycle never sees it coming.

System to system
Signal: no human in either seat

Two autonomous stacks, one intersection

  1. 01
    ScenarioTwo driverless vehicles from different operators meet at an ambiguous merge. Both behave reasonably by their own driving policy. They collide anyway.
  2. 02
    Machine evidenceTwo independent logs that must be reconciled: what each stack perceived, what each predicted the other would do, and where the two driving policies disagreed about right of way.
  3. 03
    Where liability landsBetween two software companies, with no human negligence anywhere in the record. Fault becomes a comparison of design choices — which is a product question, settled from logs neither party controls alone.
  4. 04
    What we rate onDriving-policy conservatism, prediction accuracy against real interactions, log retention and shareability, and interoperability behaviour in mixed autonomous traffic.

Why it matters: it is the end state of the shift. There is no driver anywhere in this claim, and no personal auto policy has a role in resolving it.

Find out what you're actually carrying.

30 minutes. Tell us where your vehicles run and what your current policy says. We'll give you a candid read on where your liability sits today, and whether anything is covering it. No pitch.

hello@exaalabs.com Book on calendar