Insurance AI in Kentucky Car Accident Claims

Artificial intelligence already works inside many insurance operations. It sorts documents, identifies patterns in large files, routes work, supports estimates, and flags files for closer review. Those are different jobs: a system that screens an application for coverage performs a different function from a tool that organizes records after a crash. Kentucky applies its existing insurance law to all of them.

Kentucky Law and Insurer AI

Kentucky has not created a separate set of claim rules for each algorithm. Its Department of Insurance has taken a simpler position: when an insurance company uses AI to support a consumer decision, the result still has to comply with existing insurance law. The insurance company remains responsible for its process, its data, and the outcome.

Bulletin 2024-02

The Kentucky Department of Insurance issued Bulletin 2024-02 on April 16, 2024. It addresses insurance companies that develop, acquire, or use artificial intelligence systems in the business of insurance, and it applies to decisions and actions that affect consumers, including conduct supported by an automated system.

The bulletin says those decisions must comply with applicable insurance laws and regulations. It specifically identifies unfair trade practices, unfair claims settlement practices, and unfair discrimination as areas where existing rules continue to apply, so a carrier gets no different legal standard because software contributed to a decision.

Kentucky also urges insurance companies to maintain a written program for the responsible use of AI, covering governance, risk management, internal controls, testing, data practices, and oversight. The Department may examine an insurance company’s systems, controls, testing methods, data, and outcomes during an investigation or examination.

Insurance Company Accountability

Kentucky’s bulletin places responsibility on the insurance company even when a vendor supplied the technology. An insurance company may license a model, buy a data set, or use a service that evaluates documents without owning every part of the system, and vendor involvement does not move the insurance company’s regulatory duties to someone else.

A responsible-use program should match the nature of the system and the harm an error could cause. A low-risk tool that sorts mail may need different controls from a system that influences eligibility, pricing, or claim handling, and the bulletin calls for a risk-based approach over one checklist for every use.

Kentucky Claims Handling Laws

Kentucky’s insurance rules apply to claims whether the carrier uses a spreadsheet, a specialized model, or a traditional manual process. The bulletin specifically points to unfair claims settlement practices as an existing area of law that remains relevant to AI-supported conduct.

An error does not automatically amount to insurance bad faith. The facts, the policy, the investigation, the carrier’s explanation, and the governing law all weigh in, and an automated step is one fact within that larger record.

2026 NAIC AI Pilot

The National Association of Insurance Commissioners organized a 12-state AI evaluation pilot that runs from March through September 2026. Participating regulators are testing a common evaluation tool with volunteer insurance companies across property and casualty, life, and health insurance.

The pilot is designed for regulators. It gives them a structured way to examine governance, risk controls, data, and systems that present a higher risk to consumers. The pilot creates no national claims law and no settlement formula, and it decides nothing about whether a particular carrier handled an individual file correctly.

Kentucky and the Pilot States

Kentucky is not listed among the 12 pilot states, so it is not testing the NAIC evaluation tool during the project period. The state still has its 2024 bulletin, its insurance statutes and regulations, and the Department’s authority to investigate insurance companies.

The pilot may produce a more consistent examination method across participating states, and other regulators may study or use parts of it later. The NAIC pilot summary does not say that Kentucky has committed to future adoption or that the project will create a binding national standard.

Regulatory Examination of AI Systems

A regulator examining an insurance company’s AI program may ask practical questions: who owns the system inside the company, who approved its use, which consumer decisions it can affect, what data it uses, and how the company tests for errors, unfair discrimination, and outcomes that conflict with insurance law.

Insurance Uses of AI

The NAIC’s artificial intelligence materials describe uses across insurance operations, including underwriting, pricing, fraud detection, customer service, and claims functions. The label “AI” can refer to a simple classification model, a rules-based workflow, computer vision, natural-language processing, or a more complex predictive system.

Regulators focus on the connection between the tool and the consumer outcome. A carrier should be able to identify the purpose of a system, the data it uses, the people accountable for it, and the controls that address foreseeable errors. A high-level statement that a process is “automated” answers none of those questions.

Underwriting

Underwriting concerns whether a carrier will issue or continue a policy and on what terms, which is separate from evaluating damage or injury after a collision. A May 2026 agreement between the Pennsylvania Attorney General and GEICO provides a concrete example. According to the Attorney General’s announcement, an AI-featured tool selected a new policyholder for underwriting review, inadequate communication contributed to a loss of coverage, and GEICO agreed to improvements intended to prevent confusion and unfair auto insurance cancellations.

The matter involved no Kentucky injury settlement. It did not establish that an automated system independently canceled the policy, and the agreement included no admission of wrongdoing. It showed that a regulator can examine how an automated underwriting process affects a consumer and require changes to notices and procedures.

Claims Handling

Claims handling begins after a reported loss. Insurance companies may use software to organize incoming documents, detect duplicate material, route a file, support a repair estimate, review photographs, or identify information that needs attention. Some uses affect speed without deciding the substance of the claim, and others influence what a reviewer sees first.

The primary sources for this article do not show that every carrier lets an algorithm set an injury settlement. They show widespread regulatory interest in AI systems and describe claims as one operational area where insurance companies use technology, with the level of automation varying by company, product, and task. How carriers score injury claims is covered in how insurers use AI to score injury claims.

AI Oversight Inside an Insurance Company

Kentucky’s bulletin focuses on three concrete controls: reliable data, meaningful human review, and records that explain how an AI-supported decision was reached.

Data Quality

Every automated system depends on information. A vehicle estimate may use photographs, repair data, parts prices, labor rates, mileage, condition, and comparable vehicles. A document tool may rely on text extracted from forms and records, and a routing system may use claim type, location, reported severity, or other coded fields.

Errors can enter before the model does any work. A photograph may be unclear, a vehicle trim level may be wrong, a document may be scanned badly, or a code may describe an early impression instead of a later diagnosis. Uncorrected, those problems let a system process the wrong information very efficiently.

Human Review

The phrase “human in the loop” can describe very different arrangements. One reviewer may study the source material and make an independent decision, while another sees only a score and approves it in seconds. Both processes involve a person, with very different quality of oversight.

Kentucky’s bulletin focuses on governance. An insurance company should define roles, document responsibility, train the people who use the system, and create a way to address errors, and senior management and the board should receive information appropriate to the risk the company’s AI systems present.

Decision Records

A carrier should be able to identify which system was used, which version was active, what information entered the system, and how the output moved through the business process. Without that trail, the carrier may struggle to explain why two similar files produced different outcomes.

Version history counts because models and rules change. A result produced in January may come from different logic than a result produced in July, a vendor can update a product, a carrier can alter a threshold, and a new field can be added to a workflow.

Consumer notices are another part of the record. The Pennsylvania GEICO matter centered in part on communication around underwriting review and cancellation, and a process can fail even when the internal system performs as designed if the consumer receives no clear explanation or meaningful opportunity to respond. AI in legal work is covered separately in AI in personal injury law and personal injury firms and AI.

Reviewing an AI-Supported Decision

A useful review separates industry-wide financial trends from evidence about the specific insurance company, system, and decision at issue.

2025 Industry Underwriting Results

AM Best reported that the U.S. property and casualty industry produced its strongest underwriting performance in a decade in 2025, with direct premiums written of about $1.11 trillion, an increase of roughly 5 percent from 2024, and a combined ratio of about 93.

Those reports describe industry finances. They do not analyze individual Kentucky settlements or show that AI caused the improvement, and AM Best discusses pricing, underwriting actions, risk selection, and technology among a broader set of factors. A company-level result cannot establish what happened in one consumer’s file.

Current Public Record

The public record supports several conclusions. Kentucky expects AI-supported insurance decisions to follow existing law. The NAIC is testing a regulator evaluation tool in 12 states. Pennsylvania resolved a consumer-protection investigation involving an AI-featured underwriting review, and AM Best reported strong 2025 industry results.

Some louder claims circulating online go beyond that record. Nothing in it shows that nearly every injury settlement is selected by AI, that adjusters cannot depart from a model’s range, that the 2026 pilot created a new lawsuit for consumers, or that 2025 underwriting profit is tied to a plan to suppress Kentucky settlements.

Insurance Company Process Review

A dispute involving technology still begins with the policy, the facts, and the insurance company’s stated reason. The relevant questions include whether the company used accurate information, investigated the matter, followed its own procedure, and gave a clear explanation. Technology can add another layer of evidence on top of that foundation.

A lawyer evaluating a specific Kentucky matter may need to determine whether an automated system played any role at all. A quick decision or a standardized letter does not answer that question, and the answer depends on records from the carrier and the circumstances of the file. Sam Aguiar Injury Lawyers reviews these decisions in Kentucky car accident cases.