Evolving Jurisprudence on Criminal Law and Artificial Intelligence

The intersection of artificial intelligence and criminal law has become a critical frontier for legal scholarship and practice. As AI systems grow more autonomous, courts and legislatures must grapple with questions of liability, intent and accountability that the classical criminal law did not anticipate. This article surveys the principal debates and the frameworks now emerging.

Can AI be a legal subject?

At the centre of the debate is whether an AI system can hold legal personality — the capacity to bear rights and obligations. Scholarship distinguishes between narrow (weak) AI, which operates within defined parameters, and the hypothetical strong AI possessing human-like consciousness. The prevailing view, reflected in the European Union’s approach and in Indian law, is that existing AI systems are tools — legal objects, not subjects — and that criminal responsibility for harm they cause must rest with human actors: developers, deployers and users. Narrow AI lacks the subjective consciousness on which the concepts of intention and negligence depend.

Key challenges in prosecuting AI-related harm

1. Mens rea and causation

Criminal liability ordinarily requires a guilty mind. Where an autonomous system causes harm — a chatbot that encourages self-harm, an algorithm that discriminates, an autonomous vehicle that kills — two problems arise: whether an algorithmic output can reflect intent at all, and how far the chain of causation extends across those who designed, trained, deployed and used the system. Courts have so far answered by holding companies and individuals liable in negligence or under product-liability principles for failing to implement reasonable safeguards.

2. Evidentiary hurdles

• Opacity — “black-box” models make it difficult to trace how inputs produced outputs.
• Volatility — self-learning systems may overwrite their own decision trails.
• Authentication — synthetic media and deepfakes challenge the integrity of digital evidence, placing new demands on the Bharatiya Sakshya Adhiniyam’s provisions on electronic records.

3. Jurisdiction

AI systems are built, hosted and used across borders. Servers, developers and victims may be in different countries, complicating investigation, extradition and enforcement.

Emerging legal frameworks

Human-centric liability

Approach	Description
Product liability	Manufacturers answer for defective systems, as in cases of AI-assisted medical misdiagnosis.
Negligence	Failures in the duty of care — inadequate safety filters, absent human oversight — are punished.
Accessorial liability	Humans who use AI to commit offences — automated fraud, cyber-attacks, non-consensual imagery — are charged as principals or abettors.

Specialised legislation

The EU Artificial Intelligence Act, adopted in 2024, prohibits certain practices outright and imposes graded obligations on high-risk systems. Other jurisdictions are developing algorithmic accountability and transparency requirements. In India, the Information Technology Act and its intermediary rules, the Digital Personal Data Protection Act, 2023 and the Bharatiya Nyaya Sanhita, 2023 together address much AI-enabled wrongdoing — fraud, impersonation, defamation, obscenity and the misuse of personal data — even though no AI-specific criminal statute yet exists.

Preparing for strong AI

Should genuinely autonomous systems emerge, legal personhood may need reassessment by reference to capacity for moral reasoning, understanding of legal consequences and economic independence. Models under discussion include limited “electronic personhood” akin to corporate personality, guardianship regimes with accountable human custodians, and strict-liability rules for specified AI harms irrespective of intent.

Conclusion

The jurisprudence of AI and criminal law is developing through adapted doctrine and new legislation. For now, AI remains an instrument under human control, and the law rightly looks to the humans behind it. The priorities are harmonised international standards, investment in digital forensics and the preservation of meaningful human oversight — so that society is protected from AI-enabled harm without stifling transformative innovation.