Artificial intelligence has reshaped contemporary society by assuming cognitive functions that were once performed by human beings alone. As these systems become increasingly autonomous in healthcare, transport, finance and law enforcement, questions of accountability for the harm they occasion have grown more pressing. This paper examines whether criminal liability can attach to an artificial intelligence system, and if not, upon whom it must fall. Criminal liability in the Indian tradition rests upon two elements: the wrongful act (actus reus) and the guilty mind (mens rea). Both elements presuppose a subject capable of consciousness, moral evaluation and choice. An artificial intelligence system possesses none of these attributes, and it is therefore neither a legal person nor a bearer of criminal intention. The paper addresses three questions. First, can an artificial intelligence system be recognised as a person in criminal law? Secondly, if it can, is it capable of satisfying the standards of criminal culpability? Thirdly, do the existing Indian statutes adequately govern harm caused by autonomous systems? The study is doctrinal and comparative. It analyses the Bharatiya Nyaya Sanhita, 2023, the Information Technology Act, 2000 and the Digital Personal Data Protection Act, 2023 alongside the regulatory responses of the United States, the European Union and the United Kingdom. The paper concludes that Indian law does not treat artificial intelligence as a punishable subject, that the existing cyber-law framework is inadequate to address autonomous decision-making, and that liability must be allocated among developers, deployers and users through targeted statutory amendment rather than through the creation of electronic personhood.