Are Insurers Keeping Pace with the Changing Risks of AI?

This article is by Nik Kairinos, CEO & Co-founder, RAIDS AI

As businesses weave AI more deeply into their operations, insurers are under growing pressure to understand the risks the technology creates and offer meaningful cover when things do go wrong. Deloitte estimates that the market for AI-specific insurance could grow by approximately 80% annually, reaching $4.7 billion in global premiums by 2032. Demand for these policies is already evident. A Geneva Association survey of 600 corporate insurance decision-makers found that more than 90% were interested in cover for generative AI risks, with two-thirds willing to pay at least 10% more for such protection.

Meeting that demand, however, is far from straightforward. Unlike most traditional risks, AI does not stand still. A system with a defined purpose and tightly focused parameters when a policy is written may behave very differently a few months later, particularly if the underlying technology is updated or the business changes how it is deployed.

This leaves insurers with a difficult task: accurately assessing and pricing AI-related risks that could change significantly over the lifetime of the policy.

Why AI risk changes

There are several reasons why an AI system’s risk profile can shift after an insurance policy has been issued.

First, AI models are not static. Developers may release updates, fine-tune models or introduce new capabilities that can affect how a system behaves. These changes can introduce unexpected behaviour or vulnerabilities that were not present when the system was initially assessed.

Second, businesses may change how they use AI over time. This is particularly significant with agentic AI, where systems can carry out multi-step tasks with less direct human involvement. A financial services firm, for example, might initially use AI to help employees assess loan applications, before allowing the system to make lending decisions on its own.

Giving AI greater autonomy in this way could increase the risk of discriminatory outcomes, regulatory breaches or financial losses, with errors or unexpected behaviour potentially going unnoticed until significant harm has occurred.

Third, AI risks can shift without any deliberate changes to the technology or its deployment. AI systems can behave differently when exposed to new data or unfamiliar operating conditions, affecting their performance and reliability.

As a result, an AI system’s risk profile can evolve considerably over the course of its deployment, making it difficult for insurers to accurately assess the risks associated with its use at any given point in time.

What does this mean for insurers and brokers?

For insurers, the changing nature of AI risk creates significant difficulties for underwriting and pricing. As a business’s exposure evolves, insurers risk underestimating potential liabilities, charging premiums that no longer reflect actual exposure and facing claims they hadn’t anticipated.

There are implications for coverage, too. If a business significantly changes how it uses AI after taking out a policy, questions may arise about whether its existing cover remains appropriate. This could lead to disputes over whether claims fall within the terms of the policy, potentially leaving businesses facing substantial costs.

Insurance brokers also have an important role to play. To advise clients on appropriate coverage and policy limits, they need to understand how those clients are deploying AI. However, if a client’s use of AI changes, the advice given may no longer reflect their actual exposure.

The consequences extend to claims handling. When an AI-related incident occurs, insurers may need to establish how the system was operating at the time and what caused the failure. Without reliable records of its behaviour, investigating claims and determining liability becomes considerably more difficult.

Who is responsible for keeping track?

Insurers rely heavily on information provided by policyholders to assess their exposure. However, businesses may not always recognise when changes to their AI systems have implications for their insurance cover. They might update an AI model or expand its use without appreciating how this could affect the business’s risk profile.

This makes communication between insurers, brokers and policyholders more important than ever. Insurers need to establish clear expectations around what changes should be disclosed, while brokers must help clients recognise when changes warrant a review of their cover.

Businesses, meanwhile, need processes for documenting changes to their AI systems and assessing their impact on exposure. Without this, insurers risk relying on outdated information, while policyholders could face gaps in coverage or claims disputes over undisclosed changes.

How insurers could adapt their approach The insurance industry is already responding to AI risks. Research from the Geneva Association highlights how providers are developing specialist AI insurance products, adapting existing cyber and liability policies, and exploring new underwriting practices to address emerging generative AI-related risks. But insurers must also account for how exposure changes after policies are issued.

One starting point is more detailed underwriting assessments that consider how AI is used, its level of autonomy, and the potential consequences if something goes wrong. Insurers could also introduce regular risk reviews for businesses with significant AI exposure, rather than assessing risks only at inception or renewal.

More flexible, modular insurance products could allow cover to be adjusted as AI-related risks evolve. A better understanding of these risks could also influence decisions about coverage limits, exclusions, renewal terms, and premiums.

Continuous, independent monitoring should underpin this approach by providing evidence of how AI systems behave in real-world conditions. Identifying deviations from expected behaviour could help insurers spot evolving risks and provide valuable records for claims investigations.

Insurers could also use their influence to encourage stronger AI governance among policyholders. For higher-risk applications, they could require businesses to demonstrate appropriate monitoring and governance controls as a condition of cover. Alternatively, they could offer more favourable terms to businesses with stronger safeguards in place, as they may present a lower risk than businesses without such measures.

As AI becomes more deeply embedded in business operations, insurers cannot afford to treat risk assessments as a one-off exercise. Instead, they need ongoing visibility into AI behaviour and regular reviews of businesses’ exposure to ensure their policies keep pace with evolving risks.

About alastair walker 21436 Articles
20 years experience as a journalist and magazine editor. I'm your contact for press releases, events, news and commercial opportunities at Insurance-Edge.Net

Be the first to comment

Leave a Reply

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.