# California: insurance AI rules and how they differ from the NAIC model bulletin

> California is recorded as 'Independent framework' on the NAIC adoption map, under Bulletin 2022-5. California's insurer AI guidance predates the model bulletin by eighteen months and was written for a different job. Bulletin 2022-5 restates existing anti-discrimination law and reserves the Department's examination rights; it does not set up a governance program, and the Department has issued no AI bulletin since.

- Source page: https://insureaiwire.com/states/california/
- Publication: InsureAI Wire
- Difference layer reviewed: 2026-08-18
- Measured against: naic-model-bulletin-2023-12-04 (parallel-framework)
- NAIC AI Systems Evaluation Tool pilot: participating. The trial runs from March to September 2026. Nothing in Bulletin 2022-5 refers to it. (https://content.naic.org/sites/default/files/call_materials/Pilot%20Project%20Summary.pdf)
- Method: every departure below is quoted from both instruments with a locator. Where one side has nothing to quote, the reading that established the absence is given in its place. Not legal advice.

## What is different

| Point | NAIC model | California |
|---|---|---|
| Disparate impact named | — | Named in terms, but no test is prescribed |
| Adverse action notice | Tell consumers AI is in use | Give the specific reason at the adverse action |
| Written AI programme | A written programme is expected | — |
| Third-party oversight | Due diligence plus audit rights | One due-diligence sentence, no vendor terms |
| Who it addresses | Insurers holding a certificate | Admitted, non-admitted and licensees alike |
| Examination reach | Expect to be asked at examination | Right to audit models expressly reserved |

An em dash means that instrument says nothing on the point. That is a finding, not a gap in the review.

## Every difference, with both texts

### Who and what it covers

#### Covers more insurers: California addresses non-admitted companies and licensees as well as admitted insurers. The model bulletin speaks only to companies holding a certificate of authority.

- Item id: ca-scope-non-admitted
- NAIC model bulletin (Section 1, p.1): 'to remind all Insurers that hold certificates of authority to do business in the state that decisions or actions impacting consumers that are made or supported by advanced analytical and computational technologies, including Artificial Intelligence (AI) Systems … must comply with all applicable insurance laws and regulations.'
- California (Bulletin 2022-5, header): 'TO: All Admitted and Non-Admitted Insurance Companies, Licensees, and Other Interested Parties'
- Source: Bulletin 2022-5 · California DOI, retrieved 2026-08-18 via curl + pypdf (https://www.insurance.ca.gov/0250-insurers/0300-insurers/0200-bulletins/bulletin-notices-commiss-opinion/upload/BULLETIN-2022-5-Allegations-of-Racial-Bias-and-Unfair-Discrimination-in-Marketing-Rating-Underwriting-and-Claims-Practices-by-the-Insurance-Industry.pdf)

### The written programme

#### Not required here: The whole centre of the model bulletin is missing here. California asks for no written AI program, no governance structure and no inventory. It tells insurers that existing law already binds them and that the Department may come look.

- Item id: ca-no-ais-program
- NAIC model bulletin (Section 3, p.4): 'all Insurers authorized to do business in this state are expected to develop, implement, and maintain a written program (an "AIS Program") for the responsible use of AI Systems that make, or support decisions related to regulated insurance practices.'
- Not in California. Basis for that reading: Bulletin 2022-5 sets up nothing resembling an AIS Program. Across the whole 3-page, 9,037-character text: "document" 0 hits, "monitor" 0 hits, "test" 0 hits, "inventory" 0 hits, "governance" 0 hits. pypdf with loose-whitespace re-verification, 2026-08-18.
- Source: Bulletin 2022-5 · California DOI, retrieved 2026-08-18 via curl + pypdf (https://www.insurance.ca.gov/0250-insurers/0300-insurers/0200-bulletins/bulletin-notices-commiss-opinion/upload/BULLETIN-2022-5-Allegations-of-Racial-Bias-and-Unfair-Discrimination-in-Marketing-Rating-Underwriting-and-Claims-Practices-by-the-Insurance-Industry.pdf)

### Vendors and outside data

#### Asks for less here: The model devotes a section to third-party data and systems: due diligence, audit rights, cooperation with regulators. California has one sentence of due diligence and never mentions the supplier at all.

- Item id: ca-third-party-thin
- NAIC model bulletin (§4.1, p.7): 'Due diligence and the methods employed by the Insurer to assess the third party and its data or AI Systems acquired from the third party to ensure that decisions made or supported from such AI Systems that could lead to Adverse Consumer Outcomes will meet the legal standards imposed on the Insurer itself.'
- California (Bulletin 2022-5, p.3): 'before utilizing any data collection method, fraud algorithm, rating/underwriting or marketing tool, insurers and licensees must conduct their own due diligence to ensure full compliance with all applicable laws.'
- Supporting reading: Bulletin 2022-5 never names a supplier: "vendor" 0 hits, "third part" 0 hits, "broker" 0 hits (pypdf over 9,037 characters, 2026-08-18). Its duty runs to the tool, not to whoever sold it, and it carries no contract, audit-right or cooperation terms.
- Source: Bulletin 2022-5 · California DOI, retrieved 2026-08-18 via curl + pypdf (https://www.insurance.ca.gov/0250-insurers/0300-insurers/0200-bulletins/bulletin-notices-commiss-opinion/upload/BULLETIN-2022-5-Allegations-of-Racial-Bias-and-Unfair-Discrimination-in-Marketing-Rating-Underwriting-and-Claims-Practices-by-the-Insurance-Industry.pdf)

### What you tell the customer

#### Goes further here: Where an algorithm drives a declination, a limitation or a rate increase, California says the specific reason must be given. The model bulletin asks only that consumers be told AI is in use and given appropriate access to information.

- Item id: ca-adverse-action-reasons
- NAIC model bulletin (§1.9, p.5): 'The AIS Program should include processes and procedures providing notice to impacted consumers that AI Systems are in use and provide access to appropriate levels of information based on the phase of the insurance life cycle in which the AI Systems are being used.'
- California (Bulletin 2022-5, pp.2-3): 'Additionally, when insurers use complex algorithms in a declination, limitation, premium increase, or other adverse action, the specific reason or reasons must be provided. When the reason is based upon a complex algorithm or is otherwise obscured by the technology used, a consumer cannot be confident that the actual basis for the adverse decision is lawful and justified.'
- Source: Bulletin 2022-5 · California DOI, retrieved 2026-08-18 via curl + pypdf (https://www.insurance.ca.gov/0250-insurers/0300-insurers/0200-bulletins/bulletin-notices-commiss-opinion/upload/BULLETIN-2022-5-Allegations-of-Racial-Bias-and-Unfair-Discrimination-in-Marketing-Rating-Underwriting-and-Claims-Practices-by-the-Insurance-Industry.pdf)

### What an examiner can ask for

#### Same thing, different words: Both texts land in the same place on examination: the regulator can come and look at the models themselves. California words it as a reserved right; the model bulletin words it as what an insurer should expect to be asked.

- Item id: ca-audit-reservation
- NAIC model bulletin (SECTION 4, p.7): 'Regardless of the existence or scope of a written AIS Program, in the context of an investigation or market conduct action, an Insurer can expect to be asked about its development, deployment, and use of AI Systems, or any specific Predictive Model, AI System or application and its outcomes'
- California (Bulletin 2022-5, p.3): 'The Department reserves the right to audit and examine all insurer business practices including an insurer's marketing, rating, claim, and underwriting criteria, programs, algorithms, and models. The examinations may include examinations within the scope of market conduct examinations or SIU examinations.'
- Source: Bulletin 2022-5 · California DOI, retrieved 2026-08-18 via curl + pypdf (https://www.insurance.ca.gov/0250-insurers/0300-insurers/0200-bulletins/bulletin-notices-commiss-opinion/upload/BULLETIN-2022-5-Allegations-of-Racial-Bias-and-Unfair-Discrimination-in-Marketing-Rating-Underwriting-and-Claims-Practices-by-the-Insurance-Industry.pdf)

### What the words mean

#### Extra requirement here: California names proxy discrimination and disparate impact in terms. The model bulletin uses neither word. Naming is as far as it goes, though: the bulletin prescribes no test and sets no threshold.

- Item id: ca-disparate-impact-named
- Not in the model bulletin. Basis for that reading: "disparate" 0 hits and "proxy" 0 hits in the model bulletin; its operative wording is "bias analysis and minimization" (§3.2) and "unfair discrimination" (§2.4). Conversely "test" is 0 hits in Bulletin 2022-5, so naming the concept is not the same as requiring measurement. pypdf, 2026-08-18.
- California (Bulletin 2022-5, p.2): 'A growing concern is the use of purportedly neutral individual characteristics as a proxy for prohibited characteristics that results in racial bias, unfair discrimination, or disparate impact.'
- Source: Bulletin 2022-5 · California DOI, retrieved 2026-08-18 via curl + pypdf (https://www.insurance.ca.gov/0250-insurers/0300-insurers/0200-bulletins/bulletin-notices-commiss-opinion/upload/BULLETIN-2022-5-Allegations-of-Racial-Bias-and-Unfair-Discrimination-in-Marketing-Rating-Underwriting-and-Claims-Practices-by-the-Insurance-Industry.pdf)


## What the regulator asks for

| Obligation | Who | Cadence | Due | Citation |
|---|---|---|---|---|
| There is no filing to make. What there is instead is a standing reservation: the models and the criteria behind them are examinable, through a market conduct exam or an SIU exam. | All admitted and non-admitted insurers and licensees | on-request |  | Bulletin 2022-5, p.3 |
| The specific reason for a declination, limitation or premium increase has to be given, and the fact that an algorithm produced it is not an excuse for vagueness. | Any insurer whose algorithm drives an adverse action | event-driven | When you send the decision | Bulletin 2022-5, pp.2-3 |

- Bulletin 2022-5, p.3: 'The Department reserves the right to audit and examine all insurer business practices including an insurer's marketing, rating, claim, and underwriting criteria, programs, algorithms, and models. The examinations may include examinations within the scope of market conduct examinations or SIU examinations.'
- Bulletin 2022-5, pp.2-3: 'Additionally, when insurers use complex algorithms in a declination, limitation, premium increase, or other adverse action, the specific reason or reasons must be provided.'

NAIC AI Systems Evaluation Tool pilot: participating. The trial runs from March to September 2026. Nothing in Bulletin 2022-5 refers to it. (https://content.naic.org/sites/default/files/call_materials/Pilot%20Project%20Summary.pdf)

## If you write in more than one state

California is the state where nothing is due and everything is examinable. There is no report, no attestation and no program requirement, so a carrier can look compliant right up to the day a market conduct or SIU examination asks for the model. What travels badly is the paperwork habit: a governance file built for the model bulletin answers questions California never asks, and says nothing about the one it does ask, which is why this specific consumer got this specific answer. (This paragraph is InsureAI Wire's reading, resting on: ca-no-ais-program, ca-adverse-action-reasons, ca-audit-reservation, ca-exam-on-request.)

## What InsureAI Wire has written about California

- California Bulletin 2022-5 Targets AI Bias in Insurance (https://insureaiwire.com/news/california-bulletin-2022-5/) · Bulletin 2022-5 reaches marketing, rating, underwriting, claims handling, and fraud investigation, not underwriting alone.
- California Bill Would Let Auto Insurers Use Driver Telematics with Privacy Guardrails (https://insureaiwire.com/news/california-telematics-insurance-bill-privacy/) · AB 311 would amend Proposition 103 to let auto insurers rate on telematics, but only for drivers who opt in.
- AI in Homeowners Insurance Pricing and the State Rate Review (https://insureaiwire.com/ai-in-homeowners-insurance-pricing/) · The department's complaints over aerial-imagery cancellations, and where the rate-review file picks the issue up.

## Sources read

- Bulletin 2022-5, June 30, 2022 · California DOI, retrieved 2026-08-18 via curl + pypdf (3 pages, 9,037 characters) (https://www.insurance.ca.gov/0250-insurers/0300-insurers/0200-bulletins/bulletin-notices-commiss-opinion/upload/BULLETIN-2022-5-Allegations-of-Racial-Bias-and-Unfair-Discrimination-in-Marketing-Rating-Underwriting-and-Claims-Practices-by-the-Insurance-Industry.pdf)
- Model Bulletin: Use of Artificial Intelligence Systems by Insurers, adopted December 4, 2023 · NAIC, retrieved 2026-08-18 via curl + pypdf (9 pages, 28,577 characters) (https://content.naic.org/sites/default/files/inline-files/2023-12-4%20Model%20Bulletin_Adopted_0.pdf)
- AI Systems Evaluation Tool Pilot: Pilot Project Summary · NAIC, retrieved 2026-08-18 via curl + pypdf (357,714 bytes; participant list read from p.1) (https://content.naic.org/sites/default/files/call_materials/Pilot%20Project%20Summary.pdf)

Last reviewed Aug 18, 2026.