FRANCE Trends and Developments Contributed by: Emmanuèle Lutfalla and Louis Fer, Signature Litigation AARPI
embedded in conventional policies that were never designed for autonomous systems. A classification battle may determine whether cyber, professional indemnity, media liability or product liability cover responds, making the boundary between policies a genuine source of litigation between insureds and insurers. A major AI loss event could trigger coverage denials across several lines at once, with each insurer relying on different exclusions, triggers or limits. The result could be coordinated, high-volume insurance litigation rather than a single dispute over a single policy. French law may prove more adaptable than it first appears. The liability of the guardian of a thing under Article 1242 of the Civil Code, long applied to physical objects, could potentially be extended to the operator of an AI system that causes harm, on the reasoning that the operator exercises a form of use, direction and control over the system. The revised EU directive has expanded product liability to include software, open- ing yet another avenue for claimants. Recent French legal scholarship has argued that existing doctrines of civil liability are capable of absorbing AI-related harm without requiring a sui generis regime. The question is whether the courts and the legislature will share that confidence, or whether the pace and scale of AI deployment will force more structural intervention. Conclusion Contamination by substances whose long-term effects remain uncertain. Cyber-attacks that no one can reli- ably attribute. AI systems whose outputs escape the control of the people who deployed them. What these risks have in common, from a litigation perspective, is that they generate coverage disputes which existing policy wordings were never designed to resolve.
The disputes are already here. Insurers are refusing PFAS claims on the basis of exclusions that may not survive judicial scrutiny. Policyholders are contesting war exclusions invoked against cyber-attacks whose state sponsorship cannot be proved. D&O carriers are receiving notifications for greenwashing and duty of vigilance proceedings that do not correspond to any standard insuring clause. And when a generative AI system causes serious harm, multiple insurers may simultaneously deny cover, each pointing to a dif- ferent policy boundary, leaving the insured without a response from any of them. French courts will be asked to settle these questions with tools that were designed for a different era of risk. The requirement that exclusions be formal, limited and clear gives policyholders a powerful argument against vague or overbroad carve-outs. The direct-action mechanism puts insurers in front of judges even before coverage has been agreed between insurer and insured. And the concept of abusive resistance to payment means that an insurer who unreasonably delays or refuses indemnification faces exposure beyond the policy limit. Whether these existing mechanisms prove sufficient, or whether structural responses along the lines of GAREAT become necessary for cyber, AI or PFAS, is an open question. What is already clear is that the volume, complexity and novelty of insurance cover- age disputes in France will continue to increase. For practitioners on both sides, the litigation pipeline has never looked fuller.
123 I NSURANCE L I T IGATI O N 2026
CHAMBERS.COM
Powered by FlippingBook