Defective Big Data and AI-Induced Harm: A Civil Liability Model for Kazakhstan in Comparison with the European Union and the Republic of Korea

This article examines how deficiencies in data used by artificial intelligence systems become legally relevant and how the resulting harm is allocated across a multi-actor chain. This study combines doctrinal analysis, functional comparison, a targeted narrative review, and normative modelling; its creditworthiness scenario is hypothetical. Ten analytical categories are grouped into data quality, legal origin, and governance/evidential defects and linked to legal and technical sources. Kazakhstan’s civil and procedural law is compared with bounded European Union mechanisms and Korean civil and personal information legislation. The analysis identifies existing rules on presumed fault, evidence production, obstruction of expert examination, and joint causation before defining residual gaps. The proposed model preserves ordinary remedies, develops purpose-specific warranties and protected disclosure, and confines additional presumptions to specified records and causal issues. Exceptional operator liability without proof of fault requires express legislative designation and justification of residual risk beyond existing remedies. Joint and several liability depends on joint causation, not technical participation or missing documentation alone. The key contribution of the study is a domestically grounded coordination model that distinguishes existing law, interpretation, and legislative proposals without treating every inaccurate outcome as compensable harm.