AI Incidents Are Creating New Insurance Exposures Across Multiple Lines
Artificial intelligence has moved from laboratory concept to business operation. For insurers, AI now influences underwriting decisions, claims handling, risk selection, and customer service. As adoption accelerates across industries, the risks are multiplying.
An analysis by the Verisk Core Lines Emerging Issues team found that approximately 77% of AI incidents reported in 2025 held some form of insurance impact. These incidents range from malicious actors using AI agents for phishing attacks to simple human error in system deployment.
A recent security flaw in a widely used enterprise generative AI tool demonstrates the concentration risk. The bug allowed the tool to access, read, and summarize information marked confidential-bypassing data security protocols. The vendor has issued a fix, but the incident reflects a pattern of data privacy issues that have followed large language models since their public release in late 2022.
This vulnerability highlights potential systemic liability exposure across cyber and professional liability lines. When a single platform serves much of the office software market, a single flaw can affect thousands of organizations simultaneously.
Data Centers Are Becoming a Major Construction and Risk Market
The infrastructure supporting AI adoption is growing rapidly. Data center construction spending is projected to increase 23% this year over last, with estimates suggesting these facilities could generate up to $10 billion in new premium in 2026.
But growth brings new exposures. Power failures create business interruption losses. A 2025 cloud region outage in Northern Virginia alone generated a preliminary insured loss estimate of $581 million.
U.S. data center electricity demand could reach 106 GW by 2035, according to one projection. The U.S. had about 25 GW of operating data centers just two years ago. Even if that forecast is too high, data centers could account for up to 12% of peak U.S. electricity demand by 2028.
The North American Electric Reliability Corporation warns that this surge in power demand may strain the U.S. power grid over the next several years, potentially creating elevated or high risk of energy shortfalls, brownouts, or blackouts across much of the country.
Water consumption presents another risk. The average data center uses around 300,000 gallons of water per day. Larger hyperscale facilities can use up to 5 million gallons daily to cool servers and equipment. In regions already facing water scarcity, this demand has raised public concerns about supply shortages.
Beyond power and water, data center development increases exposure across general liability as construction firms and contractors expand to serve the market.
Autonomous Vehicles Present Liability Questions Without Clear Answers
Self-driving taxis now operate in U.S. cities and municipalities. Autonomous semi-trailers are being tested on highways. These systems depend on AI to coordinate sensors, cameras, and radar equipment.
Industry experts believe advances in AI learning models could lead to more human-like driving behavior. But human-like behavior doesn't mean human-like failure modes. A study by UC Santa Cruz researchers found that embodied AI systems-including self-driving cars and drones-could be vulnerable to hijacking through misleading text injected into real-world environments, such as altered road signs.
This raises a fundamental question about accident liability. As vehicles become more autonomous, responsibility may shift from individual drivers to fleet managers or manufacturers. Auto insurance products could eventually pivot away from third-party liability toward product liability coverage.
The Real Challenge Is Interconnected Risk
The emerging problem isn't any single AI-related exposure. It's how these risks intersect. AI systems concentrate digital, physical infrastructure, and liability risks in new ways-through reliance on shared platforms, the expanding footprint of data centers, and autonomous technologies.
For insurers evaluating the next phase of AI adoption, the task requires continuous reassessment of exposures, refinement of coverage frameworks, and evaluation of how these technologies reshape loss potential.
Understanding where AI for Insurance creates exposure-and how those exposures interact-is essential to managing them. The same applies to AI for Cybersecurity Analysts, given the data privacy and system security issues already emerging across the industry.
Your membership also unlocks:
