Georgia primary voters will soon decide the next candidate for Insurance and Fire Safety Commissioner, leading to consequential decisions on artificial intelligence and other issues.
In Georgia, five Democratic candidates are vying for the nomination to run against Republican incumbent John King, who is running unopposed in his party's primary. Voting takes place May 19.
Digital Insurance posed a series of questions to all candidates. King and Democratic primary candidate Clarence Blalock did not respond. Democratic primary candidate Thomas Dean responded with his petition to Fulton County Superior Court to disqualify all other candidates (see below).
This is the second in a series of three articles compiling candidates responses, focusing on
How can AI be used to unlock value for insurance policyholders and customers in the state?
A.J. Jain: AI can help policyholders if it is used responsibly and transparently. It can speed up claims processing, identify fraud, improve customer service, help homeowners understand risk and make it easier for consumers to compare policies.
But AI should be used to serve customers, not to deny claims faster or hide behind denials or drops in coverage. As Insurance Commissioner, I would support innovation that improves affordability, access and service, but only with clear rules requiring transparency, accuracy and accountability.
DeAndre Mathis: AI can be used to unlock value for customers by using it as a tool for setting property and casualty rates strictly based on safety assessments in eliminating arbitrary and subjective pricing models that take into consideration an individual's credit scores and ZIP codes.
Keisha Waites: AI can be a powerful tool for both consumers and companies:
- Fairer, more accurate pricing: Better risk modeling can reduce cross‑subsidies and help safe, lower‑risk households pay less.
- Faster, more transparent claims: AI can speed up documentation, fraud checks and payments, reducing stress after a loss.
- Proactive risk prevention: Personalized alerts about weather, maintenance and mitigation steps can help families avoid losses altogether.
As Commissioner, I'd push for AI that is explainable, auditable and demonstrably beneficial to policyholders, not just to the bottom line.
What should be done to prevent AI from being used for insurance fraud?
A.J. Jain: Georgia needs clear standards for how insurers, agents and third-party vendors use AI for legitimate purposes, such as verifying the identity of those making claims, and tracking suspicious claims patterns. We should audit AI systems, and make sure companies have strong cybersecurity protections.
But we also need to be careful that "fraud prevention" does not become an excuse to delay legitimate claims or target consumers unfairly. Fighting fraud is important, but the burden should not fall on honest policyholders who are just trying to get paid after a devastating loss.
DeAndre Mathis: Develop technology within AI to trigger when a person is attempting to use this tool to commit insurance fraud.
Keisha Waites: AI cuts both ways — fraudsters can use it too. To prevent that:
- Stronger identity and document verification: Use AI to detect synthetic identities, deepfakes and manipulated photos or invoices.
- Data‑sharing frameworks: Encourage secure, privacy‑respecting collaboration among insurers and law enforcement to spot patterns.
- Clear standards and penalties: Set expectations that misuse of AI for fraud will be aggressively investigated and prosecuted.
I would support modern fraud units, updated statutes and technical capacity inside the agency to keep pace with emerging threats.
As a regulator, what would your concerns be about insurers'
A.J. Jain: My biggest concerns would be discrimination, lack of transparency and lack of accountability.
If insurers use AI to price policies, deny claims, or decide who gets coverage, the public must have confidence that those tools are not discriminating based on ZIP code, race, income, credit history or other unfair metrics.
As Commissioner, I would require insurers to explain how these tools are being used, prove they are fair, and remain responsible for the outcome. A company should never be able to tell a consumer, "the algorithm decided," and walk away.
DeAndre Mathis: They would have to show me the data behind the actuarial calculations that are used to set the rates. And if we find that the data contains any elements that are subjective and point to a person's credit score or ZIP code, the rate will be rejected.
Keisha Waites: As a regulator, my concerns would be:
- Bias and discrimination: AI models can unintentionally replicate or amplify unfair treatment of protected classes or communities.
- Opacity: "Black box" models make it hard for consumers — and regulators — to understand why a rate or claim decision was made.
- Data privacy and security: Massive data collection increases the risk of misuse or breach.
- Systemic risk: If many carriers rely on similar models, a single error can create market‑wide instability.
I would require model governance, independent testing, explainability and regular audits, with the power to halt or modify harmful practices.
Note on candidate Thomas Dean:
Dean replied that he could not answer the questions and referred to his 28-page petition to the Superior Court of Fulton County, Georgia. The petition seeks a stay of two rulings against Dean's previous motions, which alleged that all opposing candidates are not eligible to be elected and should be disqualified. On April 16, an administrative judge confirmed King's eligibility to run. On April 27, Secretary of State Brad Raffensperger ruled that the other Democratic candidates are eligible to run.
Incumbent commissioner John King served on an appointed basis before being elected to his first full term in 2022; Dean stated this still violates the two-term limit for the post. Dean alleged that all of his Democratic primary opponents are not eligible to run because they currently or previously worked for insurance companies or have served in roles in other branches of Georgia state government.
