On Tuesday, June 2, California voters will choose their favorites among the 11 candidates running for insurance commissioner, with the top two facing a run-off in the November 3 general election. The incumbent is term-limited; four Democrats, five Republicans, one Socialist and one independent are running.
In this second of four parts, Digital Insurance heard from seven of the candidates about the use of AI in insurance, including concerns about its increased use, value it could provide, possible fraud, aerial imagery analysis, and autonomous vehicles. The other three parts cover regulatory issues, rate increases and regulation, and climate change.
Candidates Eric Thor Aarnio (R), Steven Bradford (D), Robert Howell (R) and Sean Lee (R) did not respond to the questionnaire.
What are your concerns about insurers' plans
Ben Allen (D): AI can make insurance faster and more efficient, but it also creates serious risks. My concerns include hidden bias, unfair nonrenewals, inaccurate property assessments, opaque underwriting decisions and consumers being unable to challenge automated outcomes. Insurance decisions affect whether people can keep their homes, drive legally or recover after a disaster. If insurers use AI, regulators must be able to examine the data, assumptions, vendors and models behind those decisions. Consumers should not be denied coverage or treated unfairly because of a black-box algorithm.
Keith Davis (Ind.): My concern with AI in insurance is simple: insurance companies already look for ways to reduce risk and protect profits, and AI gives them a tool that can do that faster and on a much larger scale. If we are not careful, you could have people getting higher rates, denied claims, or dropped coverage because of an algorithm they do not understand and cannot challenge. Technology should help consumers, not create another wall between families and the protection they are paying for. At the end of the day, there still needs to be a human element in insurance.
Merritt Farren (R): AI already is a massively useful productivity tool for consumers and businesses. It can enhance and speed product development and reduce costs — two things urgently needed in insurance. Concerns about AI, including privacy, have long been with us. Those concerns need to be addressed. Consumers and businesses need protections in obtaining insurance and in the insurance coverage they receive. AI itself doesn't change that need. As commissioner, consumer and business protection will be job No. 1, even as we change our approach to regulation in a way that allows insurance providers to innovate on behalf of customers.
Jane Kim (D): AI systems in insurance worsen discrimination by using proxies like ZIP code to disadvantage low-income consumers and people of color. When these systems are proprietary, they're difficult to audit. As Commissioner, I will require insurers to disclose when AI is used. I will mandate bias audits, and hold companies accountable when automated systems produce discriminatory outcomes. We don't yet fully understand what these systems are optimizing for. If an insurer can't explain why an AI model made a decision, it can't guarantee it's making the market more stable rather than less.
Stacy Korsgaden (R): I support insurers using AI to more accurately assess and price risk, because smarter models can stabilize markets and keep coverage available. But AI must never come at the expense of personal privacy. Strong rules must safeguard sensitive data and ensure these tools are used responsibly and transparently.
Eduardo "Lalo" Vargas (Soc.): If AI is not controlled by policyholders, the technology will be used to increase premiums, deny more claims and vastly increase profits for the insurance giants. The insurers will use AI tools to cut down their customer-service workforce. In terms of processing claims, this will mean even less people-driven customer support. AI in the hands of insurers will enable an accelerated processing to a denied claim — essentially scaling up denial by default. United Healthcare's AI-driven denial model was wrong
Patrick Wolff (D): In general, technology that improves risk pricing and risk analysis is a good thing. However, there are two broader concerns with AI that require consideration. First, any time new models are adopted, it is imperative to understand the underlying assumptions so that risk can be assessed. This is especially challenging for AI, so it will require extra effort and diligence. And second, we always want to interrogate the risk variables being used to make sure there are no societal issues of fairness or equity. I expect we will be grappling with these issues for many years to come.
Are there ways that policyholders and consumers can benefit from insurers' use of AI?
Ben Allen: Yes, if it is used responsibly. AI could help speed claims processing, detect errors, improve customer service, identify mitigation opportunities and make risk assessment more precise. In disaster response, better data tools could help insurers process claims faster and deploy resources more effectively. But the benefits must flow to consumers, not just company bottom lines. If AI helps identify lower risk, that should be reflected in coverage availability, pricing and mitigation credits. The Department's role is to make sure technology improves fairness, speed and transparency.
Keith Davis: AI can absolutely help consumers, and honestly, some people in the industry are scared to admit that because they think every conversation about AI has to be negative. The truth is, if AI is used correctly, it could make claims get paid faster, reduce a lot of the back-and-forth headaches people deal with, and help catch problems before they turn into major losses. Imagine a family dealing with a fire or water loss and not having to wait weeks just to get basic answers. That matters. My issue is not with the technology itself, it's making sure it is being used to help people and not just to find smarter ways to charge more or deny coverage.
Merritt Farren: Absolutely. AI provides massively powerful productivity tools that will allow insurance companies to become more efficient and lower costs to consumers and businesses. I'll also put it to work in the Department of Insurance to create much-needed efficiency in the Department and all that it does.
Jane Kim: AI can, in theory, speed claims processing, flag fraud more efficiently and help policyholders navigate complex coverage questions. There's genuine potential for faster, more accurate damage assessments after disasters that reduce the delays that force families to wait months for payment. AI tools could also help consumers compare policies and understand what they're actually buying. AI can and should serve policyholders rather than quietly screening out or discriminating against communities of color and low-income families. I support innovation that makes insurance faster and fairer but only with transparency and strong anti-discrimination guardrails in place.
Stacy Korsgaden: Yes. Used correctly, AI can absolutely help consumers. It can sort options, explain complicated choices in plain English, and speed up service so people get clearer answers faster. The key is strong rules to protect privacy and prevent abuse.
Eduardo "Lalo" Vargas: Under a private, for-profit insurance system, there is no benefit for policyholders if insurers use AI. This is because AI will simply be used to maximize profit, not consumer benefit. However, a public insurer which does not have a profit motive could use AI to properly gauge risk without denying claims, dropping policies or jacking up rates on ordinary consumers.
Patrick Wolff: Insofar as AI improves risk pricing and/or identifies new and better ways to reduce risk, then consumers and policyholders should ultimately benefit, so long as regulators maintain a competitive market with effective oversight and good transparency.
Is the use of AI for insurance fraud a big concern in the state? How would you address this if it is an issue, or prevent it from becoming an issue?
Ben Allen: It is a growing concern. AI can be used to create false documents, manipulated images, synthetic identities or more sophisticated fraud schemes. That hurts honest policyholders by increasing costs and undermining trust. The Department should strengthen fraud detection capacity, coordinate with prosecutors and law enforcement, and work with insurers on standards for identifying AI-enabled fraud. But anti-fraud tools also need safeguards. We cannot allow fraud prevention to become an excuse for unfair claim delays or blanket suspicion of legitimate policyholders.
Keith Davis: It's a real concern. A lot of people underestimate how big this could become. AI already creates fake documents, fake damage photos, fake voices and claims that look completely real. The problem is that fraud does not just hurt insurance companies, it raises costs on honest families who are already struggling with insurance prices. At the same time, AI can also be one of the best tools to catch fraud faster. The goal should not be fearing the technology. The goal should be staying ahead of it and making sure it is used to protect honest consumers, not punish them.
Merritt Farren: Yes, it is very much an issue. We'll address it by introducing our own AI that will allow us to much more efficiently identify and eliminate fraud.
Jane Kim: AI-enabled fraud is a real concern. With any new technology, some bad actors will misuse and abuse it even when most people don't. Yet, fraud-detection systems can be weaponized to deny legitimate claims by flagging them as suspicious. Both risks are real and must be taken seriously. I will require insurers to disclose fraud-detection methodologies and ensure systems are audited for accuracy and bias. We need clear appeals processes when claims are denied or flagged. I will coordinate with the Attorney General on criminal referrals for fraud. Fighting fraud while protecting legitimate claimants is possible, but requires forward-thinking regulation.
Stacy Korsgaden: AI-driven insurance fraud is a growing risk in California, and so is the potential misuse of AI in fraud investigations. The state needs advanced analytics to detect claims and organized fraud, coupled with clear standards, documented human oversight and privacy safeguards. If AI can be used to reliably remove legitimate fraud from the system, I would support its use, because it can reduce loss costs and ultimately lower rates for consumers.
Eduardo "Lalo" Vargas: Yes, it is a major concern — but specifically fraud by the insurance companies. As discussed above, insurance companies defraud and criminally exploit their customers through denying claims and price gouging customers. AI is being used to supercharge this type of fraud. As commissioner, I will strengthen the enforcement branch of the department by hiring AI specialists who can help investigators discover where this fraud is taking place.
Patrick Wolff: I am quite sure that the use of AI will only grow by both the fraudsters and by the regulators policing against fraud. There is no easy answer to this question: it will simply require ongoing effort and diligence.
Application of AI to
Ben Allen: Aerial imagery can be useful, but it must be accurate, current and subject to meaningful review. A homeowner should not lose coverage because an image misidentifies a shadow, tree, tarp or temporary condition. If insurers use aerial imagery or AI to justify non-renewals, rate changes or underwriting decisions, consumers should receive clear notice, the basis for the decision and a real opportunity to correct errors. The Department should set standards for accuracy, documentation, vendor accountability and appeal rights.
Keith Davis: I think we are heading into dangerous territory if we let algorithms sitting in an office decide whether a family gets to keep their insurance without any real human review. Aerial imagery and AI can be useful tools, but a satellite picture does not know the story of that property, the upgrades the homeowner made, or the work they have done to reduce risk. What worries me is consumers getting higher rates or losing coverage based on technology they cannot question, cannot see and do not fully understand. If companies are going to use AI like this, there needs to be transparency, accountability and real human involvement in those decisions.
Merritt Farren: The California legislature created poorly thought-out insurance legislation. We need an insurance commissioner trained in innovative business practices and the law, who can strip down and rewrite the legislative mess directly responsible for the insurance crisis we are now suffering. The laws need to be rewritten so that consumers will understand exactly what they are signing up for and what they are entitled to. Also, to ensure that consumers can cancel with one insurer and move to another, but insurers can't cancel on them easily or on short notice. That being done, the flying of drones will be of little import.
Jane Kim: Insurers using AI-analyzed aerial imagery to justify non-renewals or premium increases without allowing policyholders to see or contest the underlying data is fundamentally unfair. Homeowners are losing coverage based on algorithmic assessments of their property they can never even review, let alone dispute. That breaks the basic promise of insurance. I support legislation requiring insurers to disclose when aerial imagery is used in coverage or pricing decisions and offer a meaningful right to appeal. AI tools need to help assess risk accurately rather than serving as an opaque mechanism for insurers to cherry-pick customers they see as more desirable.
Stacy Korsgaden: Aerial imagery and other advanced data can and should be used to rate and underwrite, but with transparency and notice to the policyholder. When technology is disclosed and used responsibly, it helps insurers accurately assess wildfire, roof and property conditions so they can properly rate the risk and keep coverage available.
Eduardo "Lalo" Vargas: I agree with the legislature's decision to ban aerial imagery to calculate risk. It is fundamentally a question of privacy. Insurers will unleash AI on all sorts of information they can collect through collaboration with other corporations that collect data — far more specific and large-scale than traditional "actuarial" methods — in order to maximize profits. As commissioner, I would create a branch in the department focused on regulating the impact on AI on the insurance industry.
Patrick Wolff: AI analysis of drone aerial imagery raises a few tricky issues. One issue is the accuracy of the imagery and its analysis: this is what AB 1559 addresses. I'm wary of using the blunt tool of legislation to impose accuracy, because a competitive market should be more accurate and flexible over time. But the 180-day limit seems reasonable in this case, and there is broad support from interested parties. Trickier issues include privacy and property rights. There are no easy answers here and it will require ongoing discussion and negotiation to strike the right balance.
Ben Allen: Autonomous vehicles raise important questions about liability, safety, data access, cybersecurity and the shift from driver responsibility to manufacturer or software responsibility. The commissioner should work closely with DMV, CPUC, local governments, insurers, consumer advocates and safety regulators to ensure coverage is adequate and claims are handled fairly. Californians should not be left in a gray area after a crash involving an autonomous vehicle. We need clear rules on responsibility, minimum coverage, incident data and consumer recourse as the technology evolves.
Keith Davis: One of the biggest mistakes is treating self-driving cars like normal vehicles with better technology. They are not. At some point, we will have accidents where the "driver" did absolutely nothing wrong, but somebody still got hurt. That completely changes the insurance conversation. The consumer should not be trapped in a legal battle between the manufacturer, software company and insurance carrier while trying to recover from an accident. If these companies want to put autonomous vehicles on California roads, then they need to accept real responsibility when the technology fails. Innovation is exciting, but accountability has to come with it.
Merritt Farren: Autonomous vehicles need to be held to the same standards as human drivers. It's that simple.
Jane Kim: Autonomous vehicles create genuine regulatory complexity. California's current insurance framework was designed for human drivers. As commissioner, I will convene insurers, AV companies, and consumer advocates to develop clear liability rules and coverage requirements for AVs, ensuring consumers are protected. California is where this technology is being deployed at scale, and our regulatory framework needs to keep up.
Stacy Korsgaden: Autonomous vehicles shift risk from individual drivers to manufacturers, software and remote operators, so we cannot rely on old auto-insurance rules. I would require clear liability standards, higher fleet-level coverage and access to AV data so crash victims are made whole quickly without quietly pushing costs onto everyone else's premiums.
Eduardo "Lalo" Vargas: When autonomous vehicles are on the road, they create different risks than human-driven vehicles. The autonomous driving models that control vehicle decision-making in autonomous cars are owned by some of the largest tech companies in the country. Any increase in risk and therefore price due to the increased usage of autonomous vehicles should be paid for by the large tech companies.
Patrick Wolff: The solution is to have as much information as possible. Companies like Waymo should know how often Remote Assistance Operators (RAOs) intervene and how long communication between AVs and RAOs takes. So long as there is sufficient information, the insurance risks should be perfectly manageable. But since mandating this information is outside the jurisdiction of a state department of insurance, it probably requires legislation.
