2026-06-18 — views

Physical AI Insurance 2026 — Waymo Driverless Fleet Liability vs Tesla FSD Supervised Driving: The AV Insurance Benchmark

Waymo as driverless operator bears full product liability. Tesla FSD supervised mode splits liability between driver and software in active litigation.

Article 199 in the Physical AI Benchmark Series — AV Insurance and Actuarial Risk

Insurance is one of the most commercially significant but least-discussed dimensions of the autonomous vehicle industry. Every commercial AV vehicle must be insured, and the premium cost per vehicle directly affects fleet economics and minimum viable fare. Waymo’s commercial driverless fleet operates under a product liability framework where Waymo as the AV operator bears primary liability — there is no human driver to assign fault to. This is a fundamentally new insurance category. Tesla’s FSD in supervised mode creates a complex liability split being actively litigated in multiple US courts.

This benchmark covers the liability framework, insurance cost implications, actuarial modeling challenges, and the regulatory trajectory for AV insurance across five structured sections.

Section 1 — Why AV Insurance Is a Commercial Inflection Point

Traditional auto insurance is built on a human-driver liability model: the human driving the vehicle is assigned primary fault for a crash; premiums are priced based on driver characteristics including age, driving history, and location. When an AV removes the human driver from the liability equation, the entire insurance framework must be rebuilt from scratch.

Three AV insurance scenarios require fundamentally different treatment:

Scenario 1 — Supervised AV (Autopilot/FSD engaged with human supervisor): A hybrid liability question — who is responsible if the human should have intervened but did not? Who was “in control” at the moment of impact? US courts are actively litigating this question across multiple Tesla FSD crash cases. Tesla’s position is that the human driver is always legally responsible because FSD requires driver supervision. Plaintiff attorneys argue that if FSD was engaged and executing the action that caused the crash, Tesla’s product liability is implicated. This question is unresolved.

Scenario 2 — Driverless AV (Waymo One with no human in any capacity): A clear product liability framework. The AV operator (Waymo) and potentially the AV system developer (also Waymo) and vehicle manufacturer are the liable parties. There is no driver to assign fault to. This is the clearest and most commercially manageable AV insurance model.

Scenario 3 — Peer-to-peer AV (Tesla Network — owner’s vehicle operating as Robotaxi when owner is not present): The most complex scenario. Is the vehicle owner liable? Tesla as the operator? The AV software developer? State AV laws vary significantly. This model has no established insurance framework and remains largely theoretical until Tesla Network launches at scale.

The commercial implication of insurance premium cost: For a Waymo commercial fleet vehicle, insurance is a per-vehicle per-year fixed cost. If insurance premiums for driverless AV are high due to actuarial uncertainty about novel risk, this increases the minimum viable fare for profitable operations. The actuarial challenge: insurers need historical crash data to price risk. AV systems have limited crash history compared to hundreds of millions of human-driver vehicle-years underlying traditional auto insurance models. This actuarial uncertainty leads to higher premiums — risk loading for uncertainty — until sufficient AV-specific crash data accumulates.

The NHTSA Standing General Order (SGO) AV crash database, established in 2021, is creating the first systematic dataset of AV crash rates. Sample sizes are still small relative to human-driver baselines, but the dataset is growing with every commercial mile operated.

Section 2 — Waymo’s Insurance Model: Product Liability for Driverless AV

Insurance Dimension	Waymo Approach	Details	Commercial Implication
Primary liability framework	Waymo operates its commercial fleet under a product liability and commercial fleet insurance framework; Waymo as AV operator bears primary liability for crashes involving its driverless vehicles; there is no human driver to assign fault to in a Waymo One vehicle	When a Waymo vehicle causes a crash, the claim goes against Waymo as both the operator and the AV system developer; traditional auto insurers and specialty AV insurers write coverage for Waymo’s fleet; the policy covers third-party bodily injury and property damage claims	Waymo’s insurance model creates a direct financial incentive to improve safety: every crash increases Waymo’s insurance cost; this aligns commercial incentives with safety improvement, which is the regulatory intent of a product liability framework
Insurance premium structure (est.)	Waymo’s commercial fleet insurance premiums are not publicly disclosed; industry estimates suggest driverless AV commercial fleet premiums range from est. $5,000–$25,000+ per vehicle per year (est.) depending on fleet size, operating geography, and historical incident rate	The premium range reflects actuarial uncertainty: limited historical data for driverless commercial AV means insurers add significant risk loading; as Waymo accumulates clean safety data — more miles, more rides, fewer incidents — premiums should decline through experience rating	Per-vehicle insurance cost is a material input to Waymo’s unit economics: at est. $10,000/year per vehicle operating 20 hours/day, insurance adds est. $1.37/operating hour; this is significant relative to est. $15–$25 per ride
Self-insurance component	Large commercial fleets often partially self-insure — assume risk for claims below a retention threshold and buy excess insurance above it; Waymo may self-insure a portion of its fleet risk given Alphabet’s substantial financial resources	Self-insurance reduces premium cost but increases Alphabet’s direct financial exposure to Waymo crash claims; Waymo’s safety record (zero fatalities in commercial driverless operations as publicly reported through mid-2026) reduces the expected claims cost of self-insurance	Alphabet’s financial strength ($100B+ cash) enables Waymo to take a self-insurance approach that would not be feasible for a standalone startup; this is a structural advantage over pure-play AV competitors that must buy full commercial coverage at actuarially uncertain premiums
Incident reporting and experience rating	Waymo’s NHTSA SGO reports and California DMV incident reports create a transparent public record of Waymo’s incident history; insurers can use this public data for experience rating	Experience rating: the cleaner Waymo’s safety record, the lower its insurance premium through actuarial experience rating; this creates a virtuous cycle: better AV safety leads to lower insurance premiums, which improves unit economics, enabling more fleet investment	Waymo’s strong safety record is not just a regulatory and public relations asset — it is a direct financial asset that reduces per-vehicle insurance cost year over year
Third-party accident claims process	When a Waymo vehicle is involved in a crash, claims go directly to Waymo as the operator/liable party rather than through a driver’s personal auto insurance; claimants deal with Waymo’s insurance or legal team rather than a human driver’s insurer	This is operationally cleaner for crash victims than traditional auto insurance: there is a well-funded corporate liable party (Waymo/Alphabet) with clear insurance rather than potentially an underinsured human driver	The driverless AV product liability model potentially provides better victim compensation than human-driver insurance: corporate liability with Alphabet’s resources behind it vs. an individual driver’s policy limits
Reinsurance market development	The reinsurance market for AV is developing; Lloyd’s of London and major reinsurers have been developing AV-specific treaty reinsurance products; liquid AV reinsurance markets reduce premium volatility and enable more competitive pricing	As the AV reinsurance market matures, primary insurers can price AV risk more efficiently; currently, limited reinsurance capacity means primary insurers must hold more capital against AV risk, reflected in higher premiums	Reinsurance market development is one of the most important commercial enablers for AV scaling: without liquid reinsurance, primary insurers are capacity-constrained on how many AV fleet vehicles they can insure, limiting fleet expansion

Section 3 — Tesla FSD’s Insurance: The Supervised Driving Liability Puzzle

Insurance Dimension	Tesla Approach	Details	Commercial Implication
Supervised FSD liability split	When Tesla FSD is engaged and a crash occurs in supervised mode, US courts and insurers are actively debating the liability split: is it primarily the human driver who failed to intervene, or primarily Tesla whose software initiated the action that led to the crash?	Multiple NHTSA and NTSB investigations of Tesla FSD crashes have examined this question; Tesla’s position is that the human driver is always legally responsible because FSD requires driver supervision; plaintiff attorneys argue that if FSD was engaged and executing the action causing the crash, Tesla’s product liability is implicated	This liability question has direct insurance pricing implications: if courts consistently hold that Tesla’s FSD is the primary liable party, Tesla’s product liability exposure increases, potentially affecting Tesla’s insurance costs and liability reserves
Consumer auto insurance pricing for FSD vehicles	Major US auto insurers have started differentiating FSD-on vs FSD-off crash rates; some insurers offer Tesla FSD vehicles at lower premiums if FSD usage data is shared; others have increased Tesla FSD premiums based on crash frequency claims	Progressive, Allstate, and other major insurers have published varying positions on Tesla FSD insurance pricing; Tesla Insurance uses real-time driving data including FSD engagement data for dynamic premium pricing; traditional insurers lack the FSD-engagement data to accurately price FSD-specific risk	Consumer auto insurance pricing for FSD is becoming a data-sharing question: insurers want FSD engagement data to price risk accurately; Tesla Insurance provides this through direct data access; traditional insurers are data-limited
Tesla Insurance (first-party insurance)	Tesla offers its own insurance product in multiple US states using real-time driving behavior data from the vehicle’s sensors to price premiums dynamically; FSD engagement and safety score data directly influence premium	Tesla’s safety score (0–100) is based on driving behavior metrics including following distance, forward collision warning frequency, hard braking frequency, aggressive turning frequency, and unsafe following frequency; drivers with higher safety scores pay lower premiums	Tesla Insurance’s real-time data advantage is unique: no traditional insurer has equivalent access to real-time vehicle sensor data for premium pricing; this creates a potential competitive moat for Tesla Insurance in the FSD insurance market
Robotaxi liability framework (Austin)	Tesla’s Austin Robotaxi launch transitions toward driverless operation; as the safety monitor is removed, Tesla’s liability model must transition from supervised-driving liability to product liability similar to Waymo’s model	The regulatory and insurance framework for Tesla’s Robotaxi in driverless mode is actively being developed; Texas AV laws (among the most permissive in the US) address AV liability; Tesla will need commercial fleet insurance for its Robotaxi fleet similar to Waymo’s model	Tesla’s Robotaxi insurance transition from consumer supervised FSD (driver primarily liable) to commercial driverless operation (Tesla as operator primarily liable) is a significant commercial and regulatory milestone
Peer-to-peer Tesla Network liability	Tesla’s envisioned Tesla Network creates the most complex AV insurance scenario: is the vehicle owner, Tesla, or the AV software liable when a crash occurs while the owner’s vehicle operates as a Robotaxi without the owner present?	US states have varying frameworks for this scenario; California AB-2866 and similar legislation attempt to address peer-to-peer AV liability; commercial fleet endorsements to personal auto policies may be required for Network vehicles	Tesla Network’s peer-to-peer liability complexity is a commercial bottleneck: if Network vehicles require specialized commercial insurance at commercial fleet rates rather than personal auto rates, the economics of Network participation change significantly for vehicle owners
NHTSA/NTSB investigation history	Tesla has been subject to multiple NHTSA Special Crash Investigations and NTSB investigations of FSD-related crashes; these examine whether FSD contributed to crashes and whether Tesla’s warnings, limitations, and user interface adequately inform drivers of FSD capabilities	Each investigation creates potential liability exposure for Tesla and informs the regulatory framework for FSD-engaged driving liability; the pattern of investigations and resulting OTA recalls establishes a de facto standard for supervised AV liability	NHTSA/NTSB investigation history is a key input to actuarial risk modeling for FSD insurance: each investigation provides data on crash circumstances, contributing factors, and whether FSD behavior was within intended design parameters

Section 4 — Actuarial Modeling: Pricing a Novel Risk

Actuarial Challenge	Waymo (Driverless)	Tesla FSD (Supervised)	Resolution Path
Historical data volume	Limited: Waymo has operated commercially since 2020; cumulative driverless commercial vehicle-miles is in the est. tens of millions (est.); statistically small relative to billions of human-driver vehicle-years	Moderate: Tesla has accumulated est. 6B+ supervised FSD miles; much larger than Waymo’s driverless commercial dataset but mixes FSD-engaged and FSD-disengaged miles; FSD-engaged-only crash rate is not separately disclosed	NHTSA SGO database is the primary source of systematic AV crash data; sample sizes are growing but still small relative to the statistical needs of actuarial pricing
Failure mode characterization	AV failure modes are systematically different from human failure modes: AV does not drink and drive, does not fall asleep, does not speed for emotional reasons; but AV may fail at novel situations, edge cases, and perception errors that humans would handle correctly	FSD’s failure modes are documented through NHTSA investigations and Tesla’s own safety data; FSD v12/v13 failure mode profile differs from earlier rule-based versions (end-to-end neural nets fail differently than rule-based systems)	Actuaries must develop AV-specific failure mode taxonomies distinct from human-driver failure modes; “sensor failure” and “edge case perception error” are new failure categories with no human-driver equivalent
Geographic risk variation	Waymo operates in geographically constrained service zones (urban/suburban, good weather, HD-mapped areas); crash risk within these zones may be lower than average human-driver crash risk, reflecting favorable selection of operating conditions	FSD operates across diverse geographies including high speed, highway, urban, rural, and poor weather; greater geographic risk diversity; harder for actuaries to model risk without disaggregated data	Waymo’s constrained operating domain is actually favorable for actuarial pricing: limited geography means limited risk scenarios and more data per scenario; Tesla’s geographic diversity creates more complex actuarial modeling needs
Time-of-day and weather variation	Waymo’s driverless operation includes night operations and some weather conditions; LIDAR provides consistent night detection; actuarial models must capture night/day risk differential but Waymo’s LIDAR performance is more consistent across lighting conditions than camera-based systems	Tesla FSD operates at all times and in all weather that human drivers face; night camera performance and rain performance create higher risk scenarios; actuarial models must capture time-of-day and weather risk factors	Camera-only systems carry higher actuarial risk loading for night and adverse weather; LIDAR-based systems have more consistent actuarial risk profiles across conditions
Regulatory uncertainty risk loading	Driverless AV faces regulatory uncertainty — permit requirements, restrictions, potential suspension of operations; insurers add risk loading for regulatory uncertainty that could halt operations or change liability framework mid-policy-period	FSD faces regulatory uncertainty including potential mandatory OTA recalls, restriction of FSD to specific geographies, and new supervised driving requirements; insurers add risk loading for regulatory uncertainty	Regulatory risk loading is a premium component that will decline as regulatory frameworks stabilize; current AV insurance premiums include significant regulatory uncertainty loading

Section 5 — AV Insurance Benchmark Scorecard

Insurance Dimension	Waymo	Tesla FSD	Edge	2028 Outlook
Liability clarity	High: driverless equals a clear product liability framework; Waymo as operator is the liable party; no driver confusion	Low: supervised FSD has active litigation over the liability split between driver and Tesla; complex and evolving	Waymo (clearer liability framework for commercial driverless operations)	Tesla Robotaxi driverless will adopt similar product liability model; liability clarity improves as Tesla moves to driverless
Insurance premium efficiency	Potentially lower long-term: clean safety record plus experience rating equals declining premiums; Alphabet self-insurance capability	Potentially higher short-term for FSD: liability uncertainty equals premium loading; Tesla Insurance mitigates via real-time data	Waymo improving with track record; Tesla Insurance creates competitive advantage for Tesla consumers	Both improve with more data; driverless AV premium efficiency improves with accumulated safety record
Data advantage for pricing	Waymo’s fleet data is proprietary; Waymo can share safety data selectively to improve insurance pricing; NHTSA SGO public data available to all insurers	Tesla Insurance has unique real-time vehicle data advantage; no other insurer has equivalent FSD engagement data; Tesla Insurance competitive moat	Tesla Insurance (unique real-time data advantage for consumer FSD pricing)	Tesla Insurance data advantage grows with fleet; potential expansion to Robotaxi fleet
Peer-to-peer liability complexity	N/A: Waymo operates a dedicated fleet; no peer-to-peer scenario	High complexity: Tesla Network peer-to-peer liability is unresolved; commercial insurance requirement for Network vehicles may be a barrier to Network adoption	Waymo (N/A; Tesla has a structural complexity challenge)	Tesla Network liability framework must be resolved for Network to launch commercially; expected 2027–2028 state-by-state legislative resolution
Actuarial data accumulation	Slower: smaller fleet; but constrained operating domain means more data per scenario; quality of data is high (driverless commercial)	Faster: 6M+ vehicles; but supervised FSD engagement data is mixed with non-FSD data; FSD-only crash rate not separately published	Tesla (faster data accumulation by volume); Waymo (higher quality data per scenario)	Both accumulate data toward actuarially stable pricing; NHTSA SGO database improves industry-wide
Overall verdict	AV insurance is a commercial bottleneck that is less visible than the technology challenges but equally important for commercial viability. Waymo’s driverless product liability model is clear and improving with its safety record — each clean year of commercial operations reduces insurance premium through experience rating. Tesla faces a transitional challenge: supervised FSD has a complex, actively litigated liability split that creates premium uncertainty, while Tesla Insurance provides a proprietary data advantage that partially mitigates this through real-time FSD engagement data pricing. The 2028 outlook: Tesla’s Robotaxi transition to driverless will shift its commercial fleet to the same product liability model as Waymo, resolving the supervised-driving liability ambiguity for its commercial fleet. Tesla Network’s peer-to-peer liability remains the most structurally complex insurance question in the AV industry — and its resolution will determine whether the Tesla Network model is commercially viable at scale.

Section 6 — About This Series

This is Article 199 in the Physical AI Benchmark Series. Previous articles in this series have covered the ramp index, the humanoid five-company race, regulation, capital, compute, sensors, unit economics, the global race, HD mapping, fleet operations, software and OTA, partnerships, competitive moats, Cybercab versus Model Y, safety data, Waymo Gen 6, Optimus manufacturing, the 2030 forecast scenarios, the investor framework, Waymo’s city expansion pipeline, the software architecture, fleet depot economics, Tesla FSD timeline history, Gen 6 vehicle transition, Waymo Uber partnership analysis, valuation and IPO analysis, Tesla Optimus humanoid ramp, supply-side economics, robotaxi fare pricing, AV weather constraints, the talent war, the regulatory calendar, the AV data flywheel, the humanoid deployment tracker, the supply chain analysis, the consumer adoption demand index, AV cybersecurity attack surfaces, accessibility for elderly and disabled users, and autonomous trucking.

This article adds the insurance and actuarial risk dimension: how liability frameworks differ between driverless operators and supervised-driving software providers, what insurance premiums cost at the fleet level, and how actuaries are beginning to price a category of risk with no historical precedent. The AV insurance market is one of the most important commercial enablers — and constraints — on the pace of autonomous vehicle deployment.