While AI offers benefits like efficiency and fraud detection, it must comply with existing insurance laws regarding fair trade practices, consumer protection, and anti-discrimination.

Summary of Delaware's AI Guidance for Insurers (Bulletin No. 148 & AI Guidance Report)

Delaware's Department of Insurance has issued Bulletin No. 148, outlining strict expectations for how insurance companies use Artificial Intelligence (AI) Systems. The key message is that while AI offers benefits like efficiency and fraud detection, it must comply with existing insurance laws regarding fair trade practices, consumer protection, and anti-discrimination. The bulletin ensures that AI does not introduce unfair discrimination, inaccuracies, or opacity in decision-making.

Key Points:

1. AI is Here to Stay in Insurance
   AI is already being used in underwriting, pricing, marketing, claims processing, and fraud detection. Delaware acknowledges its benefits but warns against risks.

2. AI Can’t Violate Insurance Laws
   Insurance companies must ensure that AI-driven decisions comply with existing state and federal laws, especially regarding unfair trade practices and discrimination.

3. Mandatory AI Governance Programs
   Every insurance carrier must create a written AI System (AIS) Program that includes:

   • AI risk management

   • Internal governance structures

   • Accountability frameworks

   • Auditing and compliance procedures

4. Third-Party AI Systems Are Not Exempt
   Insurers cannot evade responsibility by using third-party AI models. They must audit and validate AI tools from external vendors to ensure fairness and compliance.

5. Regulators Can Demand AI-Related Data
   Delaware’s Department of Insurance has the authority to request informationabout any AI system insurers use, including:

   • Data sources and bias mitigation strategies

   • AI model development processes

   • Compliance documentation

   • Evidence that AI-driven pricing and claims decisions are fair

6. Human Oversight is Essential
   AI should not make major decisions without human intervention, especially when consumer harm is possible. Companies must explain how their AI makes decisions in understandable terms.

7. AI Cannot Result in Unfair Discrimination
   AI-driven pricing, claims processing, and underwriting cannot produce unfairly discriminatory outcomes, even if it’s unintentional.

Most Surprising, Controversial, and Valuable Statements

Surprising Statements

• AI bias isn’t just a possibility; Delaware expects insurers to actively prove their AI is unbiased.

• "Model drift" is identified as a concern, meaning AI performance can deteriorate over time and unintentionally lead to biased results.

• The bulletin applies to ALL AI-driven decisions in insurance, from marketing to fraud detection, not just underwriting and claims.

Controversial Statements

• AI transparency is now a regulatory demand, not a best practice. Insurers must explain how AI models function, even if they rely on proprietary algorithms.

• Insurers are responsible for AI failures even if they buy AI from external vendors. They cannot blame third-party providers if their AI systems discriminate or cause unfair pricing.

• The state explicitly reserves the right to investigate AI use at any time and issue penalties for non-compliance, which could make AI adoption riskier for some insurers.

Most Valuable Statements

• Delaware aligns with NAIC's AI principles, showing that AI regulation in insurance is becoming standardized across states.

• Insurers must document their AI governance programs thoroughly to prepare for regulatory audits, which means compliance-ready AI models will have a competitive advantage.

• Consumer harm, not just compliance, is a core focus—the bulletin sets an expectation that insurers must prove they are protecting consumers from AI-driven unfair decisions.

Final Takeaway

Delaware is taking a proactive, strict approach to regulating AI in insurance. The state wants insurers to be fully accountable for AI-driven decisions and to provide transparency on how AI influences pricing, claims, and policy approvals. AI models must be explainable, fair, and compliant—or risk regulatory scrutiny and penalties.