<![CDATA[The rapid development of Artificial Intelligence (AI) has created new challenges in criminal law, particularly when autonomous systems perform actions that cause harm and potentially fulfil the elements of a criminal offence. Indonesia’s criminal law framework, which remains anthropocentric and limits legal subjects to humans and corporations, does not yet provide an adequate mechanism to determine criminal liability for acts committed by AI. This study aims to analyze the possibility of applying criminal liability to AI as a perpetrator and to identify the parties who may be held accountable when AI acts independently beyond direct human control. This research employs a normative juridical method by examining statutory regulations, criminal law theories, legal doctrines, and international regulatory developments concerning AI. Data were collected through literature studies on digital criminal law, modern liability concepts, and global frameworks such as the European Union Artificial Intelligence Act. The findings indicate that AI cannot fulfil the mens rea requirement under classical criminal liability theory, and therefore cannot be considered a criminal legal subject in the traditional sense. However, AI-generated actions may produce real and significant legal consequences, necessitating alternative models of liability. Three approaches are relevant: vicarious liability, which attributes responsibility to developers or operators; strict liability for high-risk AI systems; and shared liability, which distributes accountability proportionally among involved parties. The concept of electronic legal personhood may also be considered to ensure functional accountability. This study concludes that Indonesia must reform its criminal law by reconstructing the concept of legal subjects, the notion of fault, and the structure of liability in order to effectively and fairly respond to the evolving challenges posed by AI technologies. Keywords: Artificial Intelligence, criminal liability, legal subject]]>
