Following the recent Insurance Insider piece on “an age of silent exposures” in AI liability, it is clear that the market is facing a growing risk of unnoticed accumulation events. These exposures are not hypothetical. They are already present in portfolios, unpriced and unmanaged, and they originate directly from the insured’s use of AI.
What is Silent AI Risk?
Silent AI refers to the unpriced risk embedded within a portfolio that originates from an insured’s use of AI. These exposures are not explicitly recognised, underwritten or excluded. They can arise when AI systems are already in operation within products, services, supply chains or internal processes yet the associated liability landscape has not been addressed in the policy.
Because these risks are silent, they can sit undetected across multiple accounts and lines of business until a common failure mode triggers correlated claims.
The Hidden CAT Potential
Widespread deployment of the same AI model or API across multiple clients can create portfolio-wide dependencies. A flaw in a foundation model, a bias in a decision engine or a hallucination in a large language model could surface simultaneously across insureds. The result is not one isolated loss but a cascade of claims that can resemble a catastrophe-level event.
Silent AI can affect:
- Cyber through data leakage or malicious prompt exploitation
- E&O / Professional Liability through incorrect outputs driving client loss
- General Liability through defective products or harm caused by autonomous systems
- Regulatory risk through breaches of evolving AI-specific compliance frameworks
- Property when AI-dependent systems fail, triggering correlated physical losses
Silent AI in Action: Two Sector Scenarios
| Financial Services | Property |
|---|---|
| Scenario: Multiple insured banks and lenders use the same AI-driven loan approval system. A subtle bias in training data causes discriminatory decisions over 18 months. | Scenario: A widely used AI-powered building fire safety platform is deployed across commercial properties. A software update introduces a classification error, misidentifying real fires as false alarms. |
| Trigger: Regulatory investigation leads to fines for equality law breaches, followed by class action claims. | Trigger: Fires break out in multiple buildings, suppression systems fail to activate, causing widespread damage. |
| Impact: 15 insured financial institutions face average fines of £5m, settlements of £3m, and remediation costs of £2m each. | Impact: 12 insured commercial properties sustain fire damage averaging £8m each, plus £2m BI losses and £1m liability claims per property. |
| Portfolio Loss: ~£150m across E&O, regulatory liability, cyber endorsements and D&O. | Portfolio Loss: ~£132m across property, BI and liability lines. |
| Why Silent: AI model embedded in core operations but not disclosed, underwritten or priced for. | Why Silent: Fire safety system treated as an operational tool, not a potential source of correlated property loss. |
Why This Matters Now
This pattern is not new. Silent cyber taught the market that cover can exist unintentionally, creating accumulation risk without deliberate underwriting. With AI adoption accelerating, there is a narrow window to avoid repeating the mistake.
Steps for Forward-Thinking Insurers and Businesses
- Portfolio audits to identify AI in use by insureds, including indirect exposure through supply chains
- Scenario-based stress testing to quantify how common AI failures could drive correlated claims
- Clarity in wordings to explicitly address AI exposures, whether including or excluding them
- Engagement with insureds to improve disclosure of AI use and risk controls
Silent AI is not a technology problem alone. It is an underwriting and risk management challenge that requires visibility, discipline and informed decision-making. Those who act now, whether to write into the class or manage it out, will be better positioned to protect portfolios and create commercial advantage before exclusions harden and opportunity closes.