Reinsurers flag systemic risk from failures in widely used AI models
AI failures tied to popular models could trigger simultaneous losses across industries and geographic boundaries, creating aggregation risks that existing insurance products don't cover, according to a Gallagher Re report released this week.
The white paper identifies a new category of systemic exposure. Unlike natural catastrophes, which have defined geographic and temporal limits, AI-driven failures can spread rapidly through shared use of the same foundation models across sectors. A flaw in a widely adopted model could generate correlated losses simultaneously across multiple industries and jurisdictions.
This dynamic is difficult to model using traditional catastrophe approaches, creating blind spots for reinsurers managing accumulated exposure across their portfolios.
Current policies leave deployers exposed
Insurance products built for cyber, technology errors and omissions, product liability, and commercial general liability don't address many AI-specific liabilities. Losses from hallucinated outputs, discriminatory model behavior, model drift, and contaminated training data often fall outside existing coverage.
Legal and regulatory developments are shifting accountability toward organizations that deploy AI systems. Vendor contracts frequently limit manufacturer liability, leaving deployers bearing the financial burden when failures occur.
Ed Pocock, global head of cyber security at Gallagher Re, said: "AI is transforming the way businesses operate, but it also introduces a new class of risks that traditional insurance policies were never designed to address."
Insurers themselves accelerating the risk
The pressure is intensifying as insurers integrate AI into underwriting, claims processing, and customer operations. McKinsey research shows AI is now central to how insurers assess risk, set pricing, and service policies.
This broader adoption increases reliance on AI systems across the industry, raising the potential for correlated failures that could ripple through multiple insurers' portfolios simultaneously.
The Association of British Insurers has flagged AI-enabled risks including fraud and operational disruption as emerging pressures alongside climate-related losses and regulatory complexity.
Product development underway
Insurers have begun introducing standalone AI insurance products and endorsements to define coverage boundaries for both generative and non-generative AI systems. These offerings aim to work alongside existing cyber, casualty, and E&O coverage while clarifying risk allocation through governance measures and contractual approaches.
Freddie Scarratt, deputy global head of InsurTech at Gallagher Re, said: "The rapid adoption of AI has outpaced the insurance market's ability to respond to the risks it creates. By working together, insurers, reinsurers, and enterprises can close the protection gap."
The Gallagher Re framework is intended to help insurers, brokers, and reinsurers assess how AI-related risks differ from traditional exposures and how they may accumulate across portfolios. Insurance professionals managing AI-related coverage should review how AI for Insurance applications and generative AI and LLM systems affect their current risk models.
Your membership also unlocks:
