The development of intelligent systems and self-learning algorithms has fundamentally challenged traditional models of criminal attribution. In this context, criminal or harmful conduct is not necessarily the direct and simultaneous product of a human actor, but may arise from interactions among the design, training, deployment, and operation of systems possessing a degree of operational autonomy. This study examines the redefinition of mens rea in relation to algorithmic conduct and identifies criteria for attributing intent, knowledge, and volition in such circumstances. Using a descriptive-analytical method and drawing on criminal law, Islamic jurisprudential, and interdisciplinary sources, the study evaluates the capacity of traditional criminal law concepts to address these developments. The findings demonstrate that, despite their technical agency and significant role in producing harmful outcomes, intelligent systems lack free will, normative awareness, and the capacity for culpability and therefore cannot independently bear criminal responsibility. Accordingly, mens rea should be assessed primarily through the conduct and role of human actors, considering factors including effective control over the system, foreseeability of risk, the possibility of intervention and prevention, compliance with precautionary duties, and the system’s transparency and explainability. Concepts such as algorithmic causation, structural negligence, distributed liability, antecedent fault, and vicarious liability can provide useful frameworks for analyzing criminal attribution in offenses arising from self-learning systems. Consequently, redefining mens rea requires moving beyond an individual-centered model toward a multilayered, structurally oriented approach that is responsive to contemporary criminal law and emerging technological challenges effectively.