AI Risks Top Insurance Concerns as States Rush to Regulate Emerging Technologies

AI Risks Rank as Top Concern in Verisk’s Emerging Issues Bracket

In a recent industry poll by Verisk, artificial intelligence (AI) and generative AI (Gen AI) emerged as the leading concern among property/casualty insurance professionals. While climate change, microplastics, and infrastructure also made the final cut, AI’s rapid development and associated risks dominated the conversation during a June 10 webinar hosted by Verisk.

State Lawmakers Respond to AI’s Growth

AI’s swift advancement has sparked a wave of legislative activity across the United States. Many states have introduced bills aiming to regulate AI, though only a few have enacted them into law so far. Laura Panesso, Verisk’s associate vice president of government relations, noted that this reflects a broader effort to keep pace with AI's quick evolution.

The National Association of Insurance Commissioners (NAIC) has issued a bulletin outlining expectations for insurers using AI. It encourages testing for bias and discrimination in AI systems. So far, 24 jurisdictions have adopted this bulletin, while states like California, Colorado, and New York have introduced their own AI-specific regulations.

Without a federal framework, states are carving their own paths, addressing issues such as:

• Requirements for AI deployers
• Ownership rights of data used in training AI models
• Algorithmic pricing and discrimination concerns

On the generative AI front, legislation is emerging to regulate deepfakes and intimate image creation. For example, Utah’s Senate Bill 26 requires clear disclosures when generative AI is used in consumer interactions and sets liability rules under consumer protection laws.

AI’s Edge Cases and Hallucinations Pose Real Risks

Despite AI’s capabilities, it remains prone to errors—particularly in “edge cases.” These occur when AI encounters situations not well represented in its training data, increasing the risk of incorrect or harmful outcomes.

Greg Scoblete, principal on Verisk’s emerging issues team, highlighted examples from automotive AI systems. Adaptive cruise control has malfunctioned after misinterpreting unusual road scenarios, such as accelerating unexpectedly due to confusing road markings or colliding with an overturned truck because the AI didn’t recognize the obstacle.

Unlike human drivers who can instinctively adapt to unexpected events, current AI systems struggle with these rare, unfamiliar inputs. This unpredictability poses challenges for insurers assessing risk and liability.

Generative AI also exhibits “hallucinations,” where it produces incorrect outputs not directly tied to data flaws but related to how it generates responses. Legal professionals have encountered this issue, with over 120 cases reporting errors in AI-generated briefs. Stanford studies confirm that generative AI can produce significant inaccuracies in legal contexts.

Scoblete points out that the legal field isn’t unique in needing high accuracy and may soon face pressure to adopt generative AI tools despite these risks.

Legal and Liability Questions Loom

At least 11 product liability lawsuits involving generative AI have been filed. The legal system is grappling with whether traditional product liability laws for physical goods apply to AI software and virtual products.

As AI becomes embedded in more physical products like vehicles, its potential to cause property damage or personal injury increases. This intersection of AI and liability will be a critical area for insurers to monitor closely.

For insurance professionals looking to deepen their understanding of AI risks and regulatory developments, exploring targeted educational resources can be valuable. Training platforms like Complete AI Training offer courses that cover AI applications and risk considerations relevant to the insurance industry.