Loading...
Loading...
California (CA) · Insurance
Updated June 2026
California's insurance market is in a structural crisis that no other state has experienced at comparable scale. State Farm General, the largest homeowners carrier in California, announced in May 2023 that it would stop accepting new homeowners applications — then in early 2024 announced non-renewals on 72,000 existing policies concentrated in high-wildfire-risk zip codes. Allstate made a similar withdrawal from new homeowners business in the state. The California FAIR Plan, the insurer of last resort administered under the oversight of California Department of Insurance (CDI) Commissioner Ricardo Lara, saw its policy count more than double between 2020 and 2024 as admitted market carriers retreated from zip codes touched by the Camp Fire (2018, Paradise — the PG&E bankruptcy event), the Dixie Fire (2021), and the January 2025 Palisades and Eaton fires in the Los Angeles area. CDI operates under Proposition 103, a 1988 ballot initiative that created a prior-approval rate filing system with a public intervenor mechanism that can challenge rate applications — a regulatory structure that makes California one of the most constrained markets in the country for actuarially rapid repricing in response to deteriorating loss experience. Commissioner Lara's December 2023 Sustainable Insurance Strategy introduced the first major regulatory modernization in decades, including permission for carriers to use forward-looking catastrophe models (rather than only historical loss data) in rate filings — a change that directly enables AI-driven wildfire risk scoring to inform actuarially filed rates for the first time. For carriers, MGAs, and insurtech firms operating in California, AI is simultaneously the solution to underwriting accuracy and a regulatory navigation challenge that requires deep CDI engagement.
Strategic planning for AI adoption, readiness assessment, and roadmap development
Workflow automation using AI, including Make.com-style automation and RPA
Predictive models, data analysis, and ML pipeline development
Text analysis, document automation, sentiment analysis, and language processing
For three decades, Proposition 103 required California homeowners rate filings to be based exclusively on historical loss experience — which meant carriers could not use forward-looking wildfire models even as the climate-driven loss environment shifted dramatically from the historical baseline. Commissioner Lara's Sustainable Insurance Strategy, finalized in late 2023 and phased in through 2024, changed this: carriers are now permitted to include catastrophe model outputs in rate filings provided the models meet CDI's technical review standards. This regulatory shift is the single most important development in California insurance AI in years. Carriers that have already built or licensed ML wildfire risk models — scored at the property level using aerial imagery, vegetation data, and ember-transport modeling — can now move those scores into actuarially filed rates rather than using them only in underwriting decisions. Verisk's Wildfire Risk Score, CoreLogic's Wildfire Risk model, and proprietary ML systems built by larger carriers are the leading implementations. The constraint is CDI's model vetting process, which requires transparency in methodology and testing for disparate geographic impact — a process that adds 6–12 months to deployment timelines but creates a defensible regulatory record. For insurtech firms and MGAs building California wildfire models, the Sustainable Insurance Strategy filing pathway is the commercial gate that determines whether a model can support admitted market business or only surplus-lines placements.
The Camp Fire in November 2018 was the costliest wildfire in world history at the time of occurrence, generating $16.5 billion in insured losses and the bankruptcy of PG&E — whose transmission lines ignited the fire in the Sierra Nevada foothills above Paradise. The January 2025 Palisades and Eaton fires in Los Angeles County produced insured losses estimated at $30–$40 billion by early industry projections, a single-event loss that exceeded all prior California wildfire records and will drive significant AI-model recalibration. The current generation of California wildfire AI models must account for three distinct fire behavior regimes that produce different risk profiles: Sierra Nevada mixed-conifer interface fires (Camp, Dixie, Caldor), coastal chaparral fires (Tubbs, Glass, Kincade in Sonoma/Napa), and urban-wildland interface in Los Angeles County (Palisades, Eaton, Woolsey). Each regime has different ember-transport distance, rate-of-spread, and structure-ignition dynamics. Carriers and reinsurers retooling their California cat models post-Palisades are incorporating structure-level vulnerability scoring — accounting for roof type, vent screens, deck materials, and defensible-space measurements — rather than relying on geography alone. Reinsurer Swiss Re, Munich Re, and RenaissanceRe all have published updated California wildfire model revisions in response to the 2025 events, and admitted carriers are using these reinsurer models as inputs into their own CDI filing submissions.
California's represented claims rate — the percentage of claims involving an attorney — is among the highest in the nation for both auto and homeowners lines, a structural characteristic that shapes how AI claims automation must be designed. NLP-based first-notice-of-loss triage in California needs to flag attorney-involvement signals at intake and route those files to experienced adjusters rather than automated settlement tracks, because AI-driven low-offer settlements on represented California claims generate bad-faith exposure under Brandt v. Superior Court and the Insurance Fair Conduct Act. Auto insurance fraud in California is organized and sophisticated: the San Fernando Valley and East Los Angeles corridors have documented chop-shop and staged-accident networks that require ML link analysis — connecting claims to specific body shops, medical clinics, and legal offices — rather than individual-claim anomaly scoring. State Farm's California claims operation, Allstate's, and regional carriers like CSAA Insurance Group (serving AAA members in Northern California) and Amica Mutual are all deploying AI fraud analytics calibrated to California-specific fraud patterns. The Low Carbon Fuel Standard (LCFS) creates an adjacent data consideration for commercial auto carriers: fleet operators seeking LCFS credit documentation are generating telematics data streams that can double as AI-driven risk inputs for commercial auto rating, though CDI has not yet issued formal guidance on telematics-based commercial auto rating.
The December 2023 Sustainable Insurance Strategy, phased in through 2024, allows California homeowners carriers to incorporate forward-looking catastrophe model outputs in rate filings for the first time since Prop 103 passed in 1988. Previously, rates could only use historical loss experience — which systematically underpriced wildfire risk as the loss environment changed. Carriers can now file AI-driven wildfire risk scores as part of their actuarial support, subject to CDI model review. This is the regulatory change that makes AI-based wildfire underwriting viable in the admitted market, not just surplus lines.
Both carriers cited the mismatch between Prop 103's rate constraints and their actual modeled loss experience in wildfire-exposed zip codes. They could not obtain approved rates that reflected their actuarial assessment of risk, so they stopped accepting exposure they could not adequately price. The gap they left has accelerated AI adoption by the carriers that stayed — particularly FAIR Plan servicing carriers and surplus-lines MGAs — because competitive advantage now comes from precision underwriting that identifies insurable risks within the distressed pool, not broad market participation.
Prop 103 requires CDI approval before rate changes take effect for personal lines, and public intervenors can challenge filings — adding 6–18 months to the approval process for contested filings. AI-based rate factors must be disclosed in the filing and can be challenged on fairness or actuarial adequacy grounds. The Sustainable Insurance Strategy streamlined cat model filing pathways, but full algorithmic transparency is required. Carriers planning California AI-driven rate changes should budget 12–24 months from model finalization to implemented rates in the admitted market.
Three pathways exist: admitted market with CDI-filed cat model scores (newly viable post-Sustainable Insurance Strategy), surplus-lines placement via E&S carriers writing California fire risk outside Prop 103 constraints, and FAIR Plan wrap-around programs where an AI-scored companion policy covers non-fire perils. Surplus-lines MGAs like Palomar Specialty, Rancho Mesa, and several Lloyd's programs are the most AI-forward in California wildfire underwriting because they operate without Prop 103 rate constraints and can implement ML risk scores immediately.
The 2025 LA fires generated simultaneous high-volume claims across Pacific Palisades and Altadena — adjacent communities with very different building stock, income demographics, and claim complexity. Carriers deployed aerial imagery AI within 48–72 hours for initial damage triage, with EagleView and Nearmap providing post-fire imagery for automated structure-loss classification. FAIR Plan, handling a disproportionate share of the affected properties due to prior admitted-market non-renewals, relied heavily on delegated claims-handling partners with AI-assisted intake to process volume that exceeded their direct-staff capacity. Estimated total insured losses from both fires are $30–40 billion, making it the largest AI claims-management stress test California carriers have faced.
Join other experts already listed in California.