Document Type : Original Research Paper

Authors

• Fatemeh Najibzadeh Vamegh Abadi ¹
• Sayyed Mohammad Hadi Ghabooli Dorafshan ²
• Hosein Naseri Moghaddam ³

¹ PhD Student, Department of Islamic jurisprudence and principles of Islamic Law, Faculty of Theology and Islamic Studies, Ferdowsi University of Mashhad, Mashhad, Iran.

² Associate Professor, Department of Islamic jurisprudence and principles of Islamic law, Faculty of Theology and Islamic studies, Ferdowsi University of Mashhad, Mashhad, Iran.

³ Professor, Department of Islamic jurisprudence and principles of Islamic law, Faculty of Theology and Islamic studies, Ferdowsi University of Mashhad, Mashhad, Iran.

Abstract

BACKGROUND AND OBJECTIVES: The rapid deployment of artificial intelligence and the Internet of Things has reshaped the allocation of liability and the logic of insurability. Performance depends on data quality and real world operating conditions. Fault attribution among multiple actors in the chain including manufacturers, service providers, and users is difficult. Divergent national interpretations of emerging rules increase legal uncertainty and complicate risk assessment. In response, new EU frameworks, notably the Artificial Intelligence Act and the revised Product Liability regime, introduce clear definitions, evidentiary presumptions on defect and causation, and compliance duties for high risk systems. These measures can improve predictability and market discipline, although multi level implementation and interpretive variation may limit effectiveness. Against this backdrop, the study explains how these instruments affect the insurability of losses caused by intelligent systems. It shows how definitions and liability presumptions can be translated into operational metrics for underwriting, pricing, and limits of indemnity. It examines how data governance and system transparency strengthen auditability and fault attribution. It also identifies how interpretive and implementation heterogeneity constrains cover design, aggregation control, and the development of innovative insurance products.
METHODS: The study uses a legislative and analytical approach across three layers. First, it conducts a normative and interpretive reading of EU texts with emphasis on the legal definition of an AI system, the classification of risk levels, compliance duties, data governance, and the obligations of providers and users. Second, it maps these duties to technical components of insurance practice, including the management of adverse selection and moral hazard, auditability of data and models, allocation of fault, and the calibrated adjustment of deductibles and limits of liability. Third, it assesses consistency between the regulations and classic insurability criteria and develops operational scenarios for related lines of business such as technology professional liability, algorithmic product liability, and business interruption arising from failures of intelligent systems. The analysis relies on desk research and regulatory documents and uses legal analogy and law and economics reasoning.
FINDINGS: Clear definitions and compliance duties, especially for high risk systems, improve behavioral and data auditability. Presumptions on defect and causation reduce evidentiary costs and increase the predictability of liability. This enables risk based pricing and the design of contractual provisions that set information duties and help organize secondary data markets. Nonetheless, the risk based approach in the AI Act still struggles to differentiate levels and types of risk with sufficient precision. Heterogeneity in classification and the influence of non technical considerations can obscure correlations among losses and raise the likelihood of rare but severe events. In practice this increases the need for prudential capital and the use of aggregation limits in policies. Although presumptions that ease the burden of proof can streamline litigation, differences in national interpretation amplify legal heterogeneity and litigation risk, which in turn raise precautionary premium rates and broaden exclusions. The absence of fully harmonized enforcement at EU level heightens the risk of divergent standards, increases multi jurisdiction compliance costs, and complicates the harmonization of general conditions of insurance. By contrast, effective implementation of data governance duties including data quality, traceability, event logging, and full model documentation creates the basis for measurable algorithmic risks and supports risk management tools such as appropriate insurance reinsurance arrangements and performance based policy provisions. These measures enhance the capacity of insurers to assess and accept risk.
CONCLUSION: Insurance remains effective in the face of AI and IoT risks when three conditions are met. First, traceability and auditability of models and data must be translated into pre loss and post loss contractual duties. Second, risk classification must move from symbolic labels to operational measurability that uses indicators of data quality, model stability, explainability, and error rates. Third, interpretive convergence across the EU must be reinforced through governance tools that include binding guidance, a technical legal arbitration body, and minimum standard templates for contractual conditions of insurance. Otherwise, emerging risks such as error propagation across ecosystem chains and synchronous shocks generated by model updates will destabilize premium rates and constrain market capacity. The paper recommends periodic regulatory review using insurability metrics, standard templates for disclosure and event logging duties, specialist training for adjusters who handle algorithmic losses, and the development of reinsurance frameworks for accumulative risks at EU scale.

Keywords

• Artificial Intelligence
• Artificial Intelligence Act
• Civil Liability
• European Union
• Insurability

Main Subjects

• New Insurance Technologies