AI is now embedded across day-to-day operations, supporting decisions, generating content, and interacting directly with stakeholders. While the benefits are clear, recent market analysis, highlights a growing disconnect – the way AI creates risk is evolving faster than the insurance market’s ability to respond.
What is becoming increasingly evident is that AI is not simply an extension of existing technology risk. It introduces new forms of exposure that sit between operational risk, professional liability and governance accountability. These risks do not always arise from a traditional failure event such as a cyber breach or system outage. In many cases, they stem from how AI behaves in normal use, through incorrect or misleading outputs (often referred to as “hallucinations”), unintended decision outcomes, or gradual deterioration in model performance over time.
From our work with clients, these exposures are already embedded in everyday processes, recruitment and performance management, communications, advisory tools and service delivery. This is no longer theoretical risk. It is part of how organisations operate today.
A consistent theme in emerging AI litigation trends is that liability is increasingly sitting with the organisation deploying the AI, rather than the technology provider. Vendor contracts typically cap liability and limit recourse, and courts are treating AI as a tool used by the organisation rather than an independent actor. This represents a material shift in risk ownership, organisations may not control how AI models are built, but they remain accountable for how those models perform in practice.
From “Silent Cover” to Defined Positions
Historically, AI has largely sat within the concept of “silent cover”, where policies respond because losses fall within existing insuring clauses and are not explicitly excluded. This mirrors the early development of cyber risk, where policies responded to technology related losses despite not referencing cyber directly.
That position is now swiftly shifting. Insurers globally are moving to define their approach to AI risk more clearly, driven by increasing litigation and legal clarity around liability, rapid regulatory developments and the potential for systemic, portfolio-wide loss across insureds
In the US market, insurers such as Berkshire Hathaway, Chubb and Travelers have already begun introducing or approval testing broad AI exclusions across liability lines. Similarly, insights from the Lloyd’s Market Association (LMA) 2025 AI exposure survey indicate that cyber and professional indemnity classes are particularly vulnerable to AI related losses.
In practical terms, this is translating into tighter policy language and exclusions which may remove cover for:
- AI-generated content (e.g. defamation or intellectual property infringement)
- Automated or agentic decision-making
- Liabilities linked to autonomous system outputs rather than human actions.
The direction of travel is clear, greater clarity in coverage, but often narrower protection.
Fragmented Coverage and Emerging Gaps
Work by Gallagher Re highlights that existing insurance programs provide partial and fragmented responses to AI risk. Cyber policies generally respond where AI is used as an attack vector, such as deepfakes or AI-enabled phishing. Professional indemnity and liability policies may respond where AI forms part of service delivery, while D&O policies may respond to governance and oversight failures.
However, many AI related exposures fall outside traditional triggers, particularly where there is no clear breach, negligence or identifiable human error. This is most evident in areas such as:
- Hallucinations and incorrect outputs
- Algorithmic bias or discrimination (AI making unfair or biased decisions)
- Model drift and performance degradation (declining AI accuracy over time)
- Manipulation of AI systems (e.g. prompt injection, attempts to manipulate AI responses).
This creates a growing gap between exposure and insurance protection, with many organisations continuing to rely on assumed or “silent” cover.
Market Direction and Emerging Solutions
The market response is evolving, but not yet mature. New entrants and Lloyd’s-backed providers such as Armilla AI, Testudo and other MGA platforms are beginning to offer standalone AI insurance products targeting algorithmic error and model performance failure, output liability and misrepresentation and third-party litigation arising from AI use.
At the same time, established insurers (including providers such as Coalition) are developing AI related endorsements within cyber insurance, covering areas such as deepfake enabled fraud and AI driven security failures.
These developments mirror the early cyber insurance market where solutions are emerging, but coverage consistency, availability and underwriting approach vary significantly, particularly in Australia.
Implications for Clients
AI risk does not sit neatly within a single policy or function. It cuts across operations, governance, technology and communications, and therefore requires a coordinated approach.
Organisations should be mindful that reliance on “silent cover” is becoming less certain, coverage may narrow or change at renewal as insurers refine their AI position, and regulators and insurers increasingly expect clear governance and oversight of AI use.
There is also a growing expectation, particularly in a D&O context, that boards and senior leadership understand where AI is being used, how it influences decisions and outputs and where insurance coverage begins and ends.
Insurers are signalling that failures in AI governance may be viewed as foreseeable and potentially uninsured, rather than incidental.
How CRIS can Support
In response to these developments, CRIS is supporting clients to take a practical and structured approach to AI risk and insurance, including:
- Developing a clear understanding of AI use across the organisation
- Identifying key risk scenarios and potential liability exposures
- Reviewing insurance programs to assess coverage reliance and gaps
Where appropriate, CRIS can also support preparedness for AI exclusion regimes and policy changes, including engagement on affirmative cover and potential write-backs. This may include assisting with market engagement and benchmarking of standalone AI insurance solutions, such as Armilla, Testudo and Lloyd’s markets, and facilitating discussions with insurers to understand coverage appetite, exclusions and underwriting expectations.
AI is not simply introducing new risks, it is changing how liability arises and how insurance responds. Exposure is increasing. Accountability is shifting toward the user. Coverage is becoming more defined and in many cases, more restrictive. The priority for organisations is not to slow adoption, but to ensure that risk understanding, governance frameworks and insurance arrangements evolve at the same pace.
References
How AI Risk Is Covered | Coalition
Smart Systems, Blind Spots: Rethinking Insurance for the AI Era | GallagherRe
Berkshire, Chubb, and Travelers Are Removing AI Coverage
Understanding AI Exposures: AI Loss Scenarios Survey Results
Coalition Adds Deepfake Response Endorsement to its Cyber Insurance Policies Globally