Commercial auto posted a $4.9 billion underwriting loss in 2024, its 14th consecutive year of red ink (AM Best, September 2025). AI in-cab cameras deployed across 2,600-plus commercial fleets now show 73% crash rate reductions over 30 months (Samsara, October 2025). The actuarial problem: fewer than three years of equipped-fleet loss experience makes those reductions difficult to file as prospective pricing credits in any state.

The 14 consecutive loss years have pushed commercial auto carriers to examine prevention technology as a pricing variable rather than a safety certification. Direct premiums written reached $72.2 billion in 2024, a 12.1% increase from 2023 (AM Best), yet the liability sub-line combined ratio barely moved, finishing at 113. Rate increases have not kept pace with severity growth compounding at roughly 8% annually for nine consecutive years, while economic inflation over the same period averaged around 3%. From tracking ISO commercial auto rating filings across a dozen states over three consecutive filing cycles, the pattern of filed fleet technology credits reveals a consistent lag: most state-approved adjustments trail vendor deployment data by two to three years, not because regulators reject the technology, but because the actuarial documentation to support prospective credits takes time to accumulate.

The structural forces behind the loss streak are well-documented. Attorney representation rates in commercial auto liability claims have risen above 50% (Insurance Thought Leadership, 2026). Third-party litigation funding extends case duration and shifts the mix of resolved claims toward higher-severity outcomes by removing the financial pressure to settle early. Nuclear verdicts in auto accident cases accounted for 22.8% of all verdicts exceeding $10 million in 2024 (NAIC). In 2024, 135 corporate-defendant lawsuits resulted in nuclear verdicts, a 52% increase over 2023, with total nuclear verdict value reaching $31.3 billion, a 116% jump (Marathon Strategies, 2025). AM Best’s assessment is direct: “Rate increases have not kept up with increases in loss costs.” These are the inputs that cameras do not change. The frequency side is where the AI technology has produced compelling preliminary evidence.

$4.9B
2024 Underwriting Loss (14th Straight)
73%
Crash Rate Reduction (Samsara, 30 Months)
48%
Harsh Event Drop in First 6 Months
5–15%
Current Carrier Telematics Credits

Two Generations of Fleet Technology: The Behavioral Distinction

First-generation commercial fleet telematics programs, the GPS tracking and engine diagnostics platforms that expanded through the 2010s, operated on an outcome-monitoring model. They measured what happened: hard braking frequency, speeding episodes, hours-of-service compliance, engine idle time. Insurance pricing programs built on first-generation telematics granted credits based on observed behavioral scores, rewarding fleets that demonstrated better patterns over prior policy periods. The causal pathway to loss reduction was indirect; the technology recorded behavior after the risk event occurred, but it did not interrupt the sequence before a claim was generated.

AI in-cab cameras are a different intervention model. Samsara’s dual-facing camera system detects driver distraction, mobile phone use, tailgating, forward-collision proximity, and drowsiness in real time. The in-cab alert fires within seconds. A driver coaching session follows within hours or days. The near-miss that would have escalated into a $180,000 bodily injury claim was interrupted before the forward-collision sequence completed. Netradyne’s platform uses multi-directional cameras with a per-trip risk score, flagging driver-specific patterns for immediate intervention. Lytx adds AI pattern recognition trained on its database of more than 300 billion miles of driving data.

The distinction between monitoring and interruption is the foundation of the pricing argument and simultaneously the source of the actuarial credibility challenge. An AI camera program with consistent coaching does not merely correlate with safer driving; it causes behavioral change fast enough to appear in carrier loss data within 12 to 18 months of deployment. That speed is why the Samsara data shows steeper early improvement curves than first-generation telematics. It is also why the credibility problem is acute: the technology’s causal mechanism is plausible and the early frequency data is real, but the equipped-fleet loss history needed to price the credit is still accumulating at commercial scale.

Capability First-Generation Telematics AI Fleet Cameras (2022+)
Intervention model Outcome monitoring after the event Real-time behavioral interruption
Alert speed Retrospective weekly score review In-cab alert within seconds
Coaching trigger Aggregate score thresholds Event-specific session within hours
Insurance pricing model Retrospective behavioral credit Prospective prevention credit (actuarial gap)
Industry loss experience 10-plus years at scale 2 to 3 years at fleet scale
Filing credibility status Established (ISO Telematics Classification Plan) Emerging; below full credibility thresholds

The Samsara Data: 30 Months Across 2,600 Fleets

The Samsara 2025 Physical Safety Report is the most comprehensive vendor-published dataset on AI camera fleet outcomes currently available. The study covered more than 2,600 fleets worldwide, drawing on approximately 20 trillion data points processed annually, and tracked outcomes over a 30-month measurement window (Samsara, October 2025).

For fleets with 175 or more vehicles using the full AI safety solution, including dual-facing dash cams, in-cab alerts, and structured driver coaching, the findings show a consistent improvement curve. In the first six months, large fleets recorded a 48% decrease in harsh driving events and an 84% decrease in mobile phone use behind the wheel. By month 30, harsh event rates had declined 69% and mobile usage had dropped 96%. Crash rates across the same population fell 73% over the full 30-month window. For the largest fleets, those with 500 or more vehicles, harsh event reductions reached 84% by month 30 and mobile usage reductions reached 98%. The CSA Unsafe Driving score, the DOT metric most directly affecting regulatory audit exposure, improved 43% for large fleets over the same period.

The trajectory within those 30 months matters as much as the endpoint. The 48% harsh event reduction in six months represents rapid, visible behavioral change, not gradual drift. But the loss implications of that behavioral change develop on a different clock. Accident frequency in commercial auto develops over 12 to 24 months from the incident; bodily injury claims develop over three to seven years. A fleet that installs cameras in 2023 and shows behavioral improvement by mid-2024 will not have fully mature loss experience for the 2023 through 2025 accident period until 2027 through 2029 at the earliest.

The survivorship bias embedded in the vendor data compounds the credibility problem. Fleets that invest in AI camera programs at early-adopter scale tend to be larger carriers with professional driver programs, dedicated safety managers, and existing coaching infrastructure. The fleets most likely to generate catastrophic commercial auto claims, small regional carriers, owner-operators, and high-turnover operations without safety coordinators, are underrepresented in the Samsara study population. A matched-control comparison between AI-equipped and non-equipped fleets with similar pre-existing safety cultures would produce a materially different frequency reduction estimate than the longitudinal vendor analysis. The directional evidence is strong. The magnitude is not yet actuarially separable from early-adopter selection effects.

The Credibility Math: From Vendor Data to a Defensible Filing

The standard actuarial credibility framework asks how many observations are needed before experience-based rate changes are statistically defensible in a state filing. The classical full credibility standard for claim frequency, widely applied in U.S. property-casualty ratemaking, requires approximately 1,082 expected claims for a 90% probability that observed experience falls within 10% of the true underlying value. Consider the arithmetic for a representative commercial fleet. A fleet of 100 vehicles generating 12 bodily injury claims per year would require approximately 90 years to accumulate 1,082 claims on its own experience. A 500-vehicle operation generating 60 annual claims reaches that threshold in about 18 years. No single fleet arrives at full credibility independently.

Actuarial practice addresses this through Buhlmann credibility blending, weighting fleet-specific experience against the industry complement by the formula Z = n / (n + k), where n is the observed claim count and k is the ratio of expected variance within groups to variance between groups. When n is small, the industry complement dominates. For a commercial auto pricing exercise in 2026, n for the AI-equipped fleet experience is at most two to three policy years of data within any given carrier’s book, translating to limited claim counts at fleet scale. The industry complement, which does not yet reflect meaningful AI camera penetration across the broader market, carries the majority of the pricing weight. The frequency credit that survives a state filing review is necessarily smaller than the 73% reduction in vendor studies, because the Z-weight assigned to that vendor data is well below 1.0.

ISO’s Telematics Classification Plan provides the most relevant filing framework for technology-based pricing adjustments in commercial auto. The plan allows carriers to file credits and debits based on verified driver behavior data, but it requires demonstrated correlation between behavioral metrics and actual loss cost differences within the carrier’s own book, typically across multiple policy periods. A carrier that deployed AI cameras in 2023 and has two years of comparative data is on the early edge of what most state bureaus will accept as credible documentation. The programs generating approved credits in 2025 and 2026 are working from observed behavioral outcomes, not prospective prevention claims derived from vendor publications.

The actuarial work papers for any AI camera filing should document the credibility blend explicitly: the Z-weight assigned to the carrier’s own AI-equipped fleet experience, the industry complement and its source, the years of data underlying each component, and the statistical basis for the frequency adjustment selected. Filing a credit based solely on vendor-published reduction rates, without credibility discounting and the carrier’s own comparative data, is unlikely to survive review. The regulators who have approved early-round credits have done so for conservatively sized adjustments with actuarial documentation showing the carrier’s internal experience driving the selection.

Severity Counterpoint: Where Cameras Run Out of Road

The frequency argument for AI cameras is real and growing stronger. The severity argument is where the technology reaches its structural limit.

A collision that occurs in an AI-equipped fleet generates exactly the same post-accident legal exposure as one in a non-equipped fleet. A commercial vehicle bodily injury claimant with a retained plaintiff attorney, backed by litigation funding, pursuing a trucking company in a high-verdict jurisdiction will produce the same severity outcome regardless of how well the fleet’s CSA Unsafe Driving score was improving before the accident. Camera footage cuts both ways in litigation: a well-edited clip of a distraction event from six months prior, surfaced during plaintiff discovery, can become the centerpiece of a reptile-theory narrative about corporate negligence and systematic safety failures. Several commercial auto defense attorneys have noted that camera evidence improves outcomes when it exonerates the driver and creates complications when the record is mixed.

Average commercial auto liability severity grew at approximately 8% annually for nine consecutive years, compared to roughly 3% general economic inflation over the same period (Risk and Insurance, 2025). Total claims severity across commercial auto lines climbed 64% since 2015. The 49 thermonuclear verdicts exceeding $100 million recorded in 2024 represented an 81.5% increase over 2023 (Marathon Strategies, 2025). Medical inflation, vehicle repair costs elevated by ADAS components and supply-chain disruption, and the structural effects of litigation funding on case duration all push severity upward independent of fleet safety programs.

The combined loss cost arithmetic illustrates the constraint. A carrier achieving a genuine 30% frequency reduction from a mature AI camera program, starting from a commercial auto liability combined ratio of 113, sees its loss ratio component decline meaningfully, moving the combined ratio toward the high 80s, all else equal. A real improvement. But if severity continues growing at 8% annually, the loss ratio drifts upward by roughly 8 points per year. The frequency gain is absorbed within four to five years by severity escalation unless the camera program also shortens average case duration, improves fault attribution at trial, or otherwise addresses the severity side of the loss cost equation. Carriers cannot price AI cameras as a permanent structural advantage without either expecting severity trends to moderate or pairing the camera program with legal defense strategies that improve severity outcomes in equipped-fleet claims.

What Carrier Programs and ISO Allow Today

Carriers are filing AI camera credits within the constraints of available experience, and the filed figures reflect credibility reality. The Hartford’s FleetAhead program offers up to a 5% per-vehicle premium discount for qualifying telematics installations in selected states (Tooher-Ferraris, 2026). Across the broader commercial auto market, fleets that share verified driving behavior data with carriers can qualify for premium discounts of 10% to 15% (industry survey data, 2026). These figures are well below the 73% crash reduction in vendor studies. They are priced where they are because they represent actuarially credible, documented behavioral improvements rather than prospective technology credits based on vendor-published performance data.

ISO’s recent commercial auto class plan update added granularity to rating variables, including vehicle age, commercial segment, and NAICS designation, creating better segmentation infrastructure for carriers building fleet-level pricing models. The update enables more precise risk stratification by fleet type before any telematics credit is applied, which matters because the variation in commercial auto loss rates across fleet sizes and industry segments is substantial. A pricing model that applies a uniform AI camera credit across all fleet types will misallocate the benefit, giving too much credit to fleets where the technology adds little marginal safety improvement and too little to fleets where it does the most work.

The practical filing path in 2026 is not to claim the full vendor-published frequency reduction. It is to file a credibility-weighted behavioral credit, typically in the 5% to 15% range depending on the quality and duration of the carrier’s own comparative data, with documentation showing the actuarial basis and the carrier’s internal loss experience driving the selection. Carriers that started collecting AI camera loss data in 2022 and 2023 are building toward the 2027 through 2028 filing cycle, when three to four years of comparative policy data will support a materially larger, more defensible credit.

Building the Actuarial Case Now

Pricing actuaries working on commercial auto fleet programs in 2026 are managing a data-availability problem: the technology is real, the behavioral outcomes are documented, and the loss experience is insufficient for a fully credible prospective filing at vendor-claimed magnitudes. Several approaches narrow the gap.

Matched comparison studies should be running now. A well-designed internal study compares loss experience for fleets that installed AI cameras in 2022 through 2023 against comparable fleets that did not, controlling for fleet size, industry segment, prior loss history, and geographic concentration. By 2027, such a study will have three to four policy years of comparative data, approaching the threshold where credibility blending assigns meaningful weight to the fleet-AI experience rather than the industry complement. Carriers that start this data collection after the market has moved will be perpetually behind the credibility curve.

Frequency and severity should be modeled separately in any AI camera filing. The vendor evidence is concentrated on frequency; severity trends are driven by litigation dynamics, verdict inflation, and litigation funding that cameras do not affect. A combined loss cost analysis will overstate the net benefit of a camera program and understate the residual severity risk that persists regardless of fleet safety improvement. State regulators reviewing commercial auto rate filings with camera credits have increasingly requested this decomposition.

Patterns across commercial auto filing reviews suggest state regulators are receptive to technology credits when actuarial documentation is rigorous, the credit magnitude is conservative relative to vendor claims, and the carrier demonstrates its own comparative loss data rather than relying on external publications alone. Progressive’s sustained commercial auto outperformance relative to the industry traces in part to its willingness to invest in telematics data infrastructure before the pricing advantage was fully legible. The AI camera analog may produce the same dynamic: carriers with the deepest equipped-fleet experience, and the documentation to prove it, will convert vendor data into approved pricing credits several years ahead of the broader market.

The technology will not resolve commercial auto’s structural severity problem. Nuclear verdicts, litigation funding, and attorney representation rates above 50% are legal and social phenomena that no camera prevents. But frequency reduction is a real, measurable component of the loss cost equation. Carriers building the actuarial case now, rigorously and with explicit credibility weighting, will convert the frequency evidence into defensible pricing credits before their competitors do. The credibility gap is real. It is also closable on a three to four year horizon for carriers that started the work in 2023.

Further Reading

Sources