Insurers Grapple with AI Liability Risks as Coverage Gaps Emerge
Growing AI-related harms challenge traditional insurance models while creating new market opportunities.
1AI risks outpacing insurance models
The RAND Corporation reports emerging AI harms like deepfakes, privacy violations, and discriminatory decisions don't fit traditional insurance frameworks. Major insurers like W.R. Berkley have already introduced exclusions for AI-related claims in D&O, E&O, and Fiduciary Liability products. Standardized forms used in 80% of US property policies now include optional AI exclusion language. This reflects insurers' struggle to quantify novel risks from rapidly adopted enterprise AI systems.
2Litigation landscape taking shape
Over 250 US lawsuits related to AI have emerged, primarily concerning copyright but also spanning privacy, fraud, discrimination, and contract disputes. The AI Incident Database documents 713 real-world cases of AI harm across 11 categories, with misinformation (586 cases) and deepfakes (346) leading. Multiple state laws now regulate AI in areas like political ads and synthetic CSAM, creating compliance risks for businesses. This evolving legal environment complicates insurers' risk assessments.
3Market polarization on coverage
While some insurers retreat from AI risks, others see opportunity - new and existing carriers are filling coverage gaps where they believe they can price the risk. The split reflects fundamental disagreements about insurability of AI systems. RAND notes this fragmentation creates uncertainty for businesses deploying AI, potentially slowing adoption as companies weigh uninsured liabilities against innovation priorities.
4Underwriting challenges intensify
W. Robert Berkley emphasized during earnings calls that underwriters must understand AI's impact on policyholders and risk profiles. The fundamental challenge lies in pricing unpredictable, systemic risks from technologies that may evolve faster than actuarial models can adapt. Traditional loss history approaches fail with AI, requiring novel risk assessment frameworks that don't yet exist at scale.
5Corporate adoption continues despite risks
Businesses continue deploying AI systems even as insurance protections remain uncertain, suggesting many view potential benefits as outweighing unquantified liabilities. This creates tension with fiduciary duties, as executives balance innovation against potential uninsured exposures. Some companies may be self-insuring AI risks or relying on vendor indemnifications, though these approaches have limitations for catastrophic scenarios.
6Regulatory coordination needed
RAND advocates for standardized AI incident taxonomies and Coverage Notices to clarify policy terms across states. The think tank sees this as critical for developing viable insurance markets, as inconsistent regulations and claim classifications currently hinder risk pooling. Such coordination would help insurers develop appropriate products while giving businesses clearer risk management frameworks.
7Specialty to mainstream transition
RAND predicts AI coverage will evolve from niche offerings to standard policy components as risk understanding improves. This mirrors historical patterns with emerging technologies like cyber insurance, which took decades to mature. The transition period creates opportunities for brokers and reinsurers to develop innovative structures, though widespread adoption requires solving fundamental pricing challenges.
8Broader implications for AI development
Insurance market dynamics may inadvertently regulate AI advancement more effectively than legislation. If insurers broadly deem certain AI applications uninsurable, development could shift toward more controllable, explainable systems. This market pressure might achieve what ethics boards cannot - forcing alignment between commercial AI deployments and manageable risk profiles, albeit through financial rather than moral imperatives.