FT investigation reveals insurers stripping AI liability from corporate policies and introducing sublimits as courts and regulators have yet to determine where AI error liability falls
A Financial Times investigation has identified a significant and accelerating shift in the corporate insurance market: major insurers — including AIG, which has filed for regulatory permission to exclude AI-related harms from its corporate liability cover — are moving to strip AI-related losses from standard policies or cap them through 'sublimits' (provisions that limit the insurer's maximum payout for a specific category of risk to an amount below the overall policy limit).
The legal liability question at the heart of this development is unresolved: courts have not yet determined how to apportion responsibility for AI errors between the AI developer (such as the lab that trained the model), the deploying business (a law firm, bank, or insurer using the model in client-facing work), and the end user. Lawyers and insurance brokers told the FT this adjudication could take years.
Businesses had typically assumed that technology errors and omissions (E&O) insurance — policies covering losses arising from technology failures or professional mistakes — would respond to AI-related claims. The insurance market is now contradicting that assumption, leaving a growing and unquantified gap in corporate risk coverage precisely as AI deployment in professional services accelerates.
In the UK, the stakes for law firms are particularly acute. Cifas, the UK fraud prevention industry body, reported that pushed reports to a record last year, and that identity fraud in banking rose year-on-year to , with AI-powered impersonation increasingly difficult to detect. UK banks are simultaneously deploying AI as a fraud-detection tool while facing AI-enabled attacks — a dynamic that creates layered liability exposure across financial services clients. The (which classifies AI systems by risk level and imposes mandatory compliance obligations on high-risk deployments) and the are diverging in their approaches to liability allocation, creating additional complexity for multinational firms.
AI-driven fraud
444,000
10%
63,678 cases
EU AI Act
UK's pro-innovation AI framework
Why this matters
The insurance market's retreat from AI liability cover is a structural legal risk event: it forces every business deploying AI — including law firms using generative AI for drafting, document review, and legal research — to confront uninsured exposure for AI errors. The immediate client demand this creates is for AI governance policies, vendor due diligence on AI tool providers, and contractual risk allocation in technology licence agreements and AI deployment contracts. The 'why now' driver is the combination of rapid AI adoption and the absence of settled judicial authority on liability — insurers are pricing that uncertainty by stepping back from coverage rather than waiting for courts to clarify the position. For UK firms, the divergence between the EU AI Act's mandatory liability framework and the UK's lighter-touch approach creates a compliance design challenge for clients operating across both jurisdictions.
On the Ground
A trainee in a technology or risk advisory team would mark up data processing agreements and technology licence agreements to identify where AI error liability is allocated between vendor and deploying firm, draft AI governance policy documents for clients establishing internal oversight frameworks, and prepare vendor due diligence questionnaires assessing AI suppliers' liability cover and indemnity positions.
Interview prep
Soundbite
Insurers excluding AI liability before courts define it leaves every professional services firm with unquantified, uninsured AI error exposure.
Question you might get
“How should a law firm advise a client deploying a third-party AI model in client-facing legal work, and what contractual protections should it seek from the AI vendor?”
Full answer
Major insurers including AIG are filing to exclude AI-related harms from corporate liability policies, or capping cover through sublimits, before courts have determined how liability for AI errors is apportioned between developers, deployers, and users. For law firms and their clients, this creates an immediate gap: businesses that assumed their technology E&O policies covered AI-related losses are discovering they do not. The practical consequence is a surge in demand for AI governance frameworks, contractual liability allocation in deployment agreements, and vendor due diligence — all areas where commercial law firms can add significant value. This connects to the broader regulatory divergence between the EU AI Act's prescriptive liability framework and the UK's voluntary approach, which creates different compliance obligations for multinational clients and sustains cross-border advisory demand.