California's Public Wildfire Model Enters Its RFP Stage

California's Department of Insurance released a Request for Expertise on March 25, 2026 under SB 429, the first step toward a state-owned wildfire catastrophe model. University consortium proposals were due June 22, 2026.

The department spent 2025 certifying the Verisk, Karen Clark and Moody's wildfire models for rate filings. SB 429 does not ask whether catastrophe models belong in ratemaking. It asks whether the state rather than three vendors should own the one that decides who keeps a homeowners policy.

Key Takeaways

Full public access to documentation, programs, underlying data and algorithms is written into the statute's definition of the model, which is the entire point of the bill.
All three proprietary models cleared PRID review in 2025, a process that validates methodology and calibration on a sampled basis rather than re-deriving every parameter.
An October 2026 grant funds the start of development, not a finished model, so nothing reaches a rate filing before the early 2030s.
The FAIR Plan asked for 35.8% and was granted 29%, a seven-point gap that shows how much scrutiny model-driven wildfire indications already draw with no public alternative to cite.
An 85% writing commitment in wildfire-distressed areas is what carriers give in exchange for using the certified models, and it runs on those models throughout the gap years.

What SB 429 Actually Builds

SB 429, chaptered October 10, 2025 as Chapter 541, creates a Wildfire Safety and Risk Mitigation Program inside the department and directs it to fund, through a competitive grant to a university-led consortium, a public wildfire catastrophe model with full public access to its documentation, programs, underlying data and algorithms.

That access clause is the bill. The three vendor models cleared Pre-Application Required Information Determination review through a process including public comment and a technical webinar, but their fire-spread algorithms, ignition probability surfaces and vulnerability functions remain proprietary. Regulators, intervenors and competing carriers see output and a methodology description, not code.

The objection intervenors raise is procedural. A homeowner whose premium doubles because a model reclassified their parcel cannot audit the reclassification, and neither, in full, can the department that approved the model. That is a genuine limit of PRID review, and one every catastrophe modeling regime shares, including Florida's hurricane commission process.

SB 429 does not amend PRID or assert the vendor models are wrong. It builds a parallel and lets adoption settle whether transparency is worth what it costs to produce. The sequence runs from a strategy group led by Cal Poly Humboldt in May 2025, to the March 2026 Request for Expertise, to an anticipated grant in October 2026 at which point development begins. The department wants fire science, mitigation, actuarial science, data science and computational modeling in one consortium, a broader mandate than any vendor's in-house team.

A Transparent Alternative Changes What a Model Choice Must Justify

Under current practice a carrier selects one approved model, or a justified blend, and documents that choice in its filing exhibits under ASOP No. 38. That works because every approved model is proprietary to roughly the same degree, so the actuary is choosing among comparably opaque alternatives on validation studies and peer use.

A public model changes the character of the comparison rather than adding another option. If the state model produces a materially lower expected annual loss for a territory than the vendor model a carrier filed on, the carrier has to explain, in a filing an intervenor can now audit against an open alternative, why it selected the model producing the higher indicated rate. The shelter of every available model being similarly opaque stops being available.

Model	Status as of mid-2026	Transparency posture
Verisk Wildfire Model (US)	PRID review completed July 2025, first model cleared	Proprietary; methodology disclosed, algorithms confidential
Karen Clark & Company US Wildfire Reference Model v3.0	PRID review completed August 2025	Proprietary; methodology disclosed, algorithms confidential
Moody's wildfire model	PRID review completed 2025	Proprietary; methodology disclosed, algorithms confidential
California Public Wildfire Model (SB 429)	RFP stage; consortium responses due June 22, 2026, grant anticipated October 2026	Fully open by statute: data, algorithms, and code public

The numbers a new model has to reproduce are already on the record. Verisk estimated $28 billion to $35 billion of insured property loss from the January 2025 Palisades and Eaton fires, roughly $20 billion to $25 billion for Palisades and $8 billion to $10 billion for Eaton. Milliman's independent estimate ran $25.2 billion to $39.4 billion. A model that cannot reproduce severity in that range for a comparable event configuration has failed before it reaches a filing.

The rate-side number is sharper. The FAIR Plan filed for a 35.8% statewide increase in October 2025, its largest request in seven years and its first built on forward-looking catastrophe models and net cost of reinsurance rather than smoothed historical losses. The department approved 29%.

Those seven points are the practical stake. A state model that later produces a lower loss cost for the same territories does not just add competition; it hands every future intervenor an itemized figure to litigate the FAIR Plan's methodology against, and gives regulators a reference point that did not exist under an all-proprietary regime.

The Gap Years Are Not Neutral

Nothing in the timeline puts a validated public model into a rate filing before the early 2030s. The October 2026 grant starts development, and the department frames the initial phase as development and demonstration with deployment later.

The science makes that conservative rather than pessimistic. California's project is explicitly modeled on Florida's Public Hurricane Loss Model, but wildfire behavior runs on fuel load, topography and human ignition patterns that shift year to year with drought and vegetation management, against a shallower validated record than Atlantic hurricane climatology.

Meanwhile the entire Sustainable Insurance Strategy rests on the three certified models. Carriers using department-reviewed catastrophe models and net cost of reinsurance in their filings must write at least 85% of their statewide voluntary market share in wildfire-distressed ZIP codes and counties.

That commitment is what keeps the admitted market from ceding more ground to the FAIR Plan while the public model does not exist. If rate adequacy behind an 85% commitment was calculated from a model a public alternative later shows to be miscalibrated for specific high-hazard territories, the interim is not a waiting period. Coverage commitments, pricing and FAIR Plan depopulation all proceed on the assumption that today's three models are right, which is the assumption SB 429 was written because some intervenors do not accept.

The vendors are unlikely to be displaced by it in any case. The closest precedent is the CMS Hierarchical Condition Category model, mandatory for Medicare Advantage risk scoring from 2004, which did not eliminate commercial risk-scoring vendors so much as move them on top of it, selling analytics and calibration around a public substrate. California's model would arrive into a market where three proprietary models are already approved and in use, competing for adoption rather than displacing anyone by fiat.

California's Public Wildfire Model Enters Its RFP Stage

Key Takeaways

What SB 429 Actually Builds

A Transparent Alternative Changes What a Model Choice Must Justify

The Gap Years Are Not Neutral

Further Reading

Sources