2026-06-18 — views

Physical AI Insurance and Liability — Who Pays When a Robotaxi Crashes and How AV Insurance Markets Are Evolving

AV insurance runs $0.15-0.35/mile (est.); Waymo holds clear operator liability while Tesla FSD supervised liability splits with drivers.

Article 127 in the Physical AI Benchmark Series — Physical AI Insurance and Liability: Who Pays When a Robotaxi Crashes, How AV Insurance Markets Are Evolving, and Whether Liability Clarity Is the Hidden Regulatory Unlock for Scale

Insurance and liability is the hidden infrastructure layer beneath every AV commercial deployment. When a Waymo crashes into a cyclist in San Francisco, who is liable — Alphabet, Waymo, Zeekr (the vehicle manufacturer), the sensor supplier, or the city? These questions determine insurance pricing, which directly affects fleet economics (currently $0.15–0.30/mile est.), and whether the commercial model is scalable at all. This article maps the AV insurance and liability landscape as a Physical AI benchmark dimension.

All figures labeled “(est.)” are derived from public market information, analyst estimates, and company disclosures rather than verified primary data.

Section 1 — Current Liability Framework: Who Pays When an AV Crashes

The legal framework governing AV liability in the United States has evolved pragmatically at the state level. Most states that have passed AV legislation assign operator liability to the entity that deploys the vehicle when no human driver is present. The frameworks diverge significantly depending on the level of autonomy and whether a human occupant was actively supervising the system.

Scenario	Current liability assignment	Legal basis	Key uncertainty
Waymo driverless vehicle hits a cyclist (no passenger fault)	Waymo (Alphabet subsidiary) bears primary liability as vehicle operator	Most US AV laws (CA, AZ, TX) assign liability to AV operator when no human driver present	Waymo has accepted this — disclosed it carries commercial AV insurance; amount not disclosed
Tesla FSD supervised crash (driver had hands on wheel)	Traditional product liability split: driver primarily liable (supervised = driver responsible); Tesla potentially liable for FSD defect if malfunction proven	NHTSA ODI investigations ongoing; Tesla argues supervised FSD = ADAS, not autonomous	Multiple NHTSA investigations into FSD crashes; Tesla has settled some cases
Tesla FSD unsupervised crash (Austin robotaxi)	Tesla as operator; same framework as Waymo if no safety driver present	Texas AV law does not require a human in the vehicle; operator liability	No established case law yet — Austin launch will create first precedent
Shared fault (AV and human driver interaction)	Comparative fault doctrine; both parties proportionally liable	Standard tort law	Establishing AV’s “percentage of fault” requires data that may be in the AV company’s possession — creating discovery friction
Software defect causing crash	Product liability claim against AV software developer (Tesla, Waymo)	Restatement Third of Torts (Products Liability)	Software updates via OTA create ongoing liability questions — did the update that enabled this behavior exist at time of crash?
Remote operator error (Waymo assistance center)	Waymo liable as employer of remote operator	Standard employment/respondeat superior doctrine	Remote operator’s intervention or non-intervention both potentially create liability

The most legally settled scenario is the fully driverless crash: Waymo operating without a human driver is unambiguously the liable party under California, Arizona, and Texas law. This clarity is commercially important — it allows Waymo to purchase known-scope insurance policies and price that cost into fleet economics. The least settled scenario is Tesla’s Austin robotaxi launch, which will generate the first driverless product liability case law under Texas AV statute.

Section 2 — AV Insurance Market: Where It Stands

The AV insurance market as of mid-2026 is a specialty segment with limited carrier participation, elevated premiums relative to standard commercial auto, and actuarial models built on sparse data. The market is evolving rapidly as safety records accumulate.

Insurance category	Current state	Key players	Cost benchmark (est.)	Trend
Commercial AV fleet insurance (Waymo, robotaxi operators)	Specialty market; limited carriers; Lloyd’s of London and specialty reinsurers most active	Lloyd’s, Munich Re, Swiss Re, Markel, AIG specialty	$0.15-0.35/mile (est.) — 2-4x standard commercial auto	Declining as safety record builds; Waymo’s 50M+ driverless miles without fatal crash is primary actuarial driver
ADAS/L2+ consumer insurance (Tesla FSD)	Handled within standard personal auto; FSD incidents trigger standard auto claims initially	State Farm, Progressive, Allstate, Geico	Negligible premium surcharge for FSD capability (~$50-200/yr est.)	Tesla’s crash rate data feeds actuarial models; FSD users may qualify for lower rates if data shows improvement
AV product liability (software defect)	Umbrella corporate coverage and E&O (Errors and Omissions) for AV companies	Specialty brokers; Waymo and Tesla carry but do not disclose	Not disclosed; Alphabet/Tesla balance sheet backstops	As incidents accumulate, AV-specific E&O market is emerging
Cargo/delivery AV insurance	Aurora (trucking), Nuro (delivery) carry commercial cargo and AV combination coverage	Specialty carriers	Higher per-incident exposure (cargo value and AV fleet value)	Aurora’s Texas commercial launch creating new actuarial dataset for trucking AV
Reinsurance capacity	Reinsurers taking AV exposure but with aggregate limits per AV company	Munich Re, Swiss Re, Hannover Re most active	Not disclosed	Catastrophic scenario (multi-vehicle AV failure) stress-tests reinsurance capacity

The $0.15-0.35/mile (est.) commercial AV insurance premium is one of the largest variable cost items in the Waymo fleet economics model. For context, standard commercial auto for a human driver typically runs $0.05-0.10/mile (est.) — making AV insurance 2-4x more expensive per mile. As Waymo’s 50M+ driverless mile safety record accumulates, actuarial uncertainty decreases and premiums should decline toward standard commercial auto rates over a 3-5 year horizon (est.).

Section 3 — State-by-State Liability Framework

The absence of federal AV legislation means that liability frameworks, insurance minimums, and permit requirements vary dramatically by state. This creates regulatory arbitrage: AV companies deploy first in the most permissive states, accumulate safety data, and use that record to unlock more restrictive markets.

State	AV liability framework	Driverless permit requirement	Insurance requirement	AV-friendliness
California	SB 1398 (2023): AV operator liable when no human driver; CA DMV driverless permit required; reporting mandated	Yes (CA DMV Driverless Deployment permit)	Commercial AV insurance required; minimum $5M per occurrence (est.)	Medium — permits exist but scrutiny high after Cruise incident
Arizona	Executive Order 2015-09 (Gov. Ducey): operator as driver; most permissive early framework; no specific AV legislation required	No state permit required (county/city may vary)	Standard commercial auto minimum and AV endorsement	Very high — first Waymo commercial market; minimal regulatory friction
Texas	HB 1736 (2017): AV operator = driver; no human required; no state permit required	No state permit required	Standard commercial auto minimum	Very high — Austin Waymo and Tesla robotaxi; Aurora trucking
Nevada	NRS 482A: comprehensive AV framework; first state to legalize AV (2011); operator liability	State permit required	$5M minimum coverage for driverless AV	High
Florida	CS/SB 932 (2023): updated AV framework; operator liability; driverless permitted	No permit required	Standard commercial auto	High — Miami potential target
New York	No comprehensive AV law; NYC TLC regulations apply; most restrictive environment	Effectively no permits available in NYC	N/A	Very low — regulatory barrier remains highest in US
Federal (NHTSA)	No federal AV law passed (SELF DRIVE Act stalled 2018); NHTSA uses existing authority for recalls and ODI investigations; voluntary guidance only	N/A — federal framework defers to states	N/A — federal minimum insurance laws apply	Medium — NHTSA has investigation authority but no permit power

The California/$5M minimum insurance requirement (est.) versus Texas/Arizona standard commercial auto minimum represents the widest insurance cost differential in the US AV market. Waymo’s Phoenix and Austin operations — its fastest-growing markets — benefit from lower insurance minimums. California, despite being Waymo’s longest-running and highest-volume market, carries the highest regulatory and insurance cost burden.

Section 4 — Insurance Cost Trajectory: What Makes It Fall

The AV insurance market will normalize as the actuarial data gap closes. Several specific drivers will compress premiums toward standard commercial auto rates over the 2026-2030 window.

Driver of cost reduction	Waymo path	Tesla path	Timeline (est.)
Safety record accumulation	50M+ driverless miles without fatal crash = primary actuarial data point; each additional mile lowers actuarial uncertainty	Supervised FSD miles feed separate actuarial model; Austin driverless miles will be most valuable data	2026-2028: Waymo rates should normalize as 100M+ driverless miles accumulate
Standardized data sharing	Waymo mandated crash reporting in CA; creates actuarial dataset	NHTSA SG reporting for FSD; less granular than CA mandate	Federal standardized AV crash data standard would accelerate actuarial model development
Reinsurance market development	More reinsurers entering = more capacity = lower primary rates	Same	2026-2028 as Aurora/Waymo commercial deployments generate clean 2-3yr loss history
State minimum coverage clarity	AZ/TX low minimums; CA $5M est. creates arbitrage (Waymo operating most in AZ/TX = lower insurance cost)	Texas low minimum = favorable for Austin launch	States without minimums = lower short-term cost but higher catastrophic risk for operators
AV-specific actuarial tables	No industry-standard AV actuarial table exists; carriers develop internal models	Same	ISO (Insurance Services Office) developing AV actuarial standards; est. 2027-2028 commercial release

The most important single actuarial event would be publication of ISO AV actuarial tables — the industry-standard pricing tool used by all US personal and commercial auto insurers. ISO developing AV-specific tables (est. 2027-2028) would trigger a market-wide repricing of AV insurance, likely downward for operators with clean safety records like Waymo, and upward for any operator accumulating incidents. Until ISO tables exist, every insurer is running proprietary models with limited data.

Section 5 — Liability and Insurance Benchmark Scorecard

Dimension	Waymo	Tesla	Edge
Current liability clarity	Clear: driverless operator = Waymo; accepted by Alphabet	Supervised: unclear split; unsupervised Austin: operator = Tesla (pending precedent)	Waymo — clearer framework in current operations
Insurance cost (est.)	~$0.15-0.35/mile (est.) — commercial AV fleet	Austin supervised: standard personal auto (lower); driverless: commercial AV rate (TBD)	Tesla — lower today (supervised = personal auto rates); similar when driverless
Safety record as actuarial asset	50M+ driverless miles = most valuable AV actuarial dataset in US	5-6B supervised miles = large but noisier dataset (human interventions add variance)	Waymo — driverless miles are cleaner actuarial signal
State exposure	Primarily CA (higher req) and AZ/TX (lower req)	Austin TX (low req) = favorable	Tesla — lower state insurance minimums
Catastrophic event risk	Multi-vehicle failure scenario = Alphabet backstop	Multi-vehicle failure scenario = Tesla balance sheet	Comparable — both have deep-pocket backstops
Long-term insurance cost trajectory	Declining as driverless miles compound	Declining if Austin driverless proves safe	Both declining; Waymo earlier on curve

The liability and insurance benchmark reveals an underappreciated dynamic in Physical AI scale economics: liability clarity is a commercial enabler. Waymo’s acceptance of operator liability — and the insurance market’s pricing of that liability at $0.15-0.35/mile (est.) — is actually a positive signal. It means the liability question is resolved, the insurance market is functioning, and the cost is knowable. The unknown is Tesla’s Austin driverless operation, which will face the same commercial AV insurance rates as Waymo once it operates without a safety driver, and which lacks Waymo’s accumulated actuarial safety record to negotiate favorable premiums.

The hidden regulatory unlock thesis holds: if federal AV legislation were to pass and standardize liability assignment (operator = liable when driverless) and mandate crash reporting, the resulting actuarial certainty would accelerate insurance market normalization and meaningfully reduce the $0.15-0.35/mile (est.) premium burden — improving AV fleet economics across the industry.

Note: All figures labeled “(est.)” are derived from public market information, analyst estimates, industry reporting, and company investor relations materials as of mid-2026. This article does not constitute investment advice.