Newsroom

If You Cannot Interrogate It, You Cannot Insure It

Willis published its latest Risk & Resilience Review this week. It is worth reading carefully, because a tier-one global broker has put something in writing that the market has largely been avoiding.

Newsroom 29 May 2026

Willis published its latest Risk & Resilience Review this week. It is worth reading carefully, because a tier-one global broker has put something in writing that the market has largely been avoiding.

The finding is not that AI is risky in some abstract sense. It is more specific than that: organisations are already relying on AI systems they cannot fully interrogate, placing trust in outputs that are not being challenged. That gap between what a system produces and what an organisation can actually explain about it is, in Willis's assessment, no longer just an operational concern. It has become a governance, liability and insurability challenge.

The implications for the reinsurance and insurance market are direct. Willis notes that the market is beginning to diverge. Some carriers are continuing to rely on traditional policy wording and assumptions about AI exposure that are silent, unexamined, and increasingly untenable. Others are introducing affirmative AI cover and tying underwriting requirements to governance and control frameworks. In plain terms: how well you can account for your AI systems is starting to affect whether and how you can insure them.

This is a structural shift, not a trend. When underwriting criteria begin to reflect AI governance posture, the organisations that have invested in provable, auditable AI infrastructure are in a fundamentally different position from those that have not. The gap between them will widen as claims emerge and precedent is set.

Willis's Chief AI Officer put it directly: many organisations are moving forward without fully understanding the systems they rely on. That creates a dangerous gap between innovation and oversight.

The answer to that gap is not better documentation or stronger policy language. It is infrastructure that makes AI reasoning transparent, traceable, and verifiable by design. That is a different class of solution from what most organisations currently have in place, and the window for making that distinction a competitive advantage is narrowing.

Source: https://www.wtwco.com/en-au/insights/2026/05/risk-and-resilience-review-ai-in-action-the-road-to-responsible-adoption