Mind Without A Body - Reconsidering Criminal Liability In The Age Of AI

Author: LawSphereX

Category: Criminal Law

Published: 2026-08-23T20:37:12.960153+00:00

Updated: 2026-08-26T17:10:27.16+00:00

MIND WITHOUT A BODY: RECONSIDERING CRIMINAL

LIABILITY IN THE AGE OF AI

Palak Phulwani & Preesha Kapoor, SVKM’s NMIMS Kirit P. Mehta School of Law

ABSTRACT

The excessive use of AI into decision making processes has left a wide gap in the notions of agency, responsibility and intent within the criminal justice system. AI systems are very much capable of independent action due to rapid evolution of this system, they challenge the foundational principles of criminal law, especially the elements of crime which are mens rea and actus reus. Therefore this paper examines this very complex question of liability for the crimes committed by AI, arguing that the current legal framework remains inadequate to address such emerging realities. Through a comparative analysis of international models, the study highlights the rapidly increasing “responsibility gap” between human oversight and machine autonomy.

The conclusion to this paper is until AI is given a legal status, liability must lay with the human controllers who include the programmers, deployers or user based on the model of responsibility. It further suggests the integration of specialised AI liability statute in india to fill the gap of accountability, ethical governance and justice in this era of advanced technology and intelligent machines. Keywords: Artificial Intelligence (AI), AI criminal liability, Autonomous systems, Mens rea, Responsibility gap

I.

INTRODUCTION

AI has transitioned to a transformative force which is now constantly shaping contemporary legal, social, economic realities. So, AI is a machine that is currently ruling the world in various sectors of the economy since it can give predictions, recommendations and influence decisions. Therefore, this is a very capable system which can sometimes be independent of human oversight because it can very smartly act, learn and adapt without immediate human interference. Therefore, it creates some gap in the traditional legal doctrines which is rooted in human intention and control. In this century, AI is dominating domains like finance, healthcare, transportation, law enforcement and much more, leaving space for instances of harm caused by this autonomous decision- maker.

Self-driving vehicles have been a part of various fatal accidents raising questions about who is at fault, manufacturer, programmer or user. Ai chatbots have generated misinformation which puts a huge question mark to the limits of algorithmic responsibility. Excessive news about fraud and defamation is now heard due to the deepfake technologies. Market manipulation has become very easy by algorithmic trading systems. Each example gives a direction to the same dilemma: should a system which can act without human intention yet produce harm, should be held criminally liable? Liability in criminal law which is guided through actus reus and mens rea because of the existence of consciousness but AI on the other hand operates without consciousness, emotion and subjective awareness.

It may execute harmful actions without the intention in the human sense contradicting the attribution of the criminal guilt concept. But the absence of guilt does not free the society of the need for accountability. Without someone being held responsible, victims would be left remediless and justice could not be upheld. Therefore, this paper investigates the central problem: Reconsidering Criminal Liability in the Age of AI where legal framework is insufficient and extremely lacking. II.

UNDERSTANDING AI AND ITS ROLE IN CRIME

Artificial Intelligence (AI) has become an important part of our daily life; it’s increasing influence on everything from education to financial markets and criminal investigations is a reflection of the same. As more autonomy is gained by AI, this raises the legal concern of criminal liability of AI, can AI be held criminally liable? And can it even commit crimes? To

address such concerns it is imperative to understand the levels of AI autonomy and nature of crimes that can be linked to AI activity.

A. LEVELS OF AI AUTONOMY

AI can be broadly categorized into four levels, on the basis of the degree of human involvement i. Assisted or tool-based ai: These refer to simper tools which are used completely by human control, humans are the ultimate authority and practice decision making. example includes Grammarly, through which only AI assistance is offered. ii. Partially autonomous systems: These types of systems execute tasks required but still have dependence on human foresight for the same. This often includes customer service chatbots1. iii. Conditionally autonomous or self-learning systems: Such type of systems showcase a more balanced level of AI autonomy and degree of human involvement, these systems apply their machine learning to improve data and adapt to new data.

An example would be self driving cars, as they adapt and adjust as per real time conditions such as traffic, road signals, road conditions without relying on human input on a real time basis2. iv. Fully autonomous systems: These types of systems refer to the ones that in which AI enjoys full autonomy with very limited human control. AI makes all the complex decisions independent from human control for example, drones used in financial and military functions. Such systems project the unpredictable nature of AI and raises the question of liability.

1 Hallevy, G.. (2013). When Robots kill: Artificial intelligence under criminal law. 1-244.

2 Mukherjee, Anirban & Chang, Hannah. (2025). Agentic AI: Expanding the Algorithmic Frontier of Creative Problem Solving. 10.48550/arXiv.2502.00289.

B. HOW AI CAN COMMIT CRIMES

AI lacks human cognitive skills and human conscious which primarily is responsible for the formation of mens rea, however AI can be used as a medium or an instrument to commit a crime. AI can cause harm which can be equivalent to criminal conduct. For example AI bots can manipulate the stock market leading to financial crimes, AI bots and malware can bypass data privacy systems leading to a breach of privacy and an increase in the rate of cybercrimes. Recently the development of deepfake technology, which is a product of AI produces defamatory content, AI can cause physical crimes as well, autonomous vehicles or robots can cause injuries or even death.

C. THE RESPONSIBILITY GAP

The concept of responsibility gap3 refers to the gap that is created due to the question of liability, wherein there is unidentifiable human fault and harm is caused by autonomous systems. This creates a lacuna in law. The current legal system struggles in assigning blame when the behaviors of AI diverges from human expectation. The European Parliament’s 2017 Resolution on Civil Law Rules on Robotics also acknowledged this gap, suggesting the introduction of ‘electronic personhood’ for advanced AI systems to ensure legal continuity. III.

THE CHALLENGE OF CRIMINAL LIABILITY

A crime refers to a wrongful act which is punishable in nature, not every wrongful act raises

criminal liability. For establishing culpability of a criminal act two essentials are required, which are actus reus and mens rea. Actus reus refers to the physical act and mens rea refers to the intent to commit the wrongful act, together they both establish criminal liability4.

A. AI AS ACTUS REUS

Even though AI lacks mens rea, it fulfills the actus reus requirement as it perform tasks, it depicts that the presence of actus reus without the presence of mens rea and without 3 Matthias, A. The responsibility gap: Ascribing responsibility for the actions of learning automata. Ethics Inf Technol 6, 175–183 (2004). https://doi.org/10.1007/s10676-004-3422-1 4 Horder, J. & Ashworth, A. (2019) Ashworth’s principles of criminal law. Ninth edition / Jeremy Horder. Oxford, United Kingdom: Oxford University Press.

any human agency creates a legal challenge. For example deepfake technology produces defamatory content which can cause harm to an individual’s reputation leading to defamation without any human intent.

B. AI AND MENS REA

However, in the modern world, with the emergence of Artificial Intelligence (AI) systems which are capable to cause harm and injury, independent from human intervention, raises the question whether such an entity can be held criminally liable. With the constant developments in AI, it can perform various complex tasks and make decisions as well but it lacks human consciousness and awareness, hence this results in a lack of mens rea, with the absence of the mental element, criminal liability will nor arise. This undermines the legal framework which relies on both essentials of actus reus and mens rea to be fulfilled to result in criminal liability. For example, a self-driving car performs all it’s actions independent from human agency, when an action results in an injury actus reus exists but the absence of intention results in no criminal liability.

C. HUMAN INVOLVEMENT AND THE ATTRIBUTION PROBLEM

In various cases which involve AI related harms, It can be seen that multiple human agents are involved as well such as users, programmers and operators, Hence determining which entity should bear the responsibility becomes difficult. Users often operate AI as per their understanding this may lead to an unintended action to occur. Progammers are often unaware as to how an AI system will behave, ever act of AI cannot be predicted. Corporations that own the AI system can be held vicariously liable, but the current legal system states criminal liability can only be attributed to legal persons and natural persons. As, previously mentioned a responsibility gap is created.

D. Doctrinal Challenges: AI and Legal Personhood

There exists a major obstacle in the current legal framework is as to the legal status of AI. Only natural persons and legal persons can be held criminally liable, AI fits into neither of the categories hence there is ambiguity in the current legal system as the liability of AI is undetermined. AI can neither be categorized as a natural person as it lacks human conscious nor as a legal person as it lacks judicial recognition.The

European Parliament’s Resolution on Civil Law Rules on Robotics (2017) has recommended the creation of an “electronic personhood” for categorising the highly autonomous AI systems in an attempt to ensure legal responsibility. IV.

EXISTING LEGAL FRAMEWORK

Legal Framework in India The current legal system prevalent in India is undeveloped in nature as it fails to address AI related criminal liability. The Bharatiya Nyaya Sanhita (BNS), 2023 is India’s current criminal code, while it has introduced certain changes and new elements to keep up with the modern era, it still lacks in addressing crimes involving AI. BNS dosen’t consist of provisions specific to crimes committed by AI systems, AI is not even recognized as a legal person to attribute any criminal liability to it directly. The Information Technology Act, 2000 addresses crimes such as cybercrime, hacking, and electronic offences hence it finds human actors liable for crimes committed due to their misuse of technology. It doesn’t address AI systems that act autonomously and cause injury, these acts fall outside it’s ambit.

Similarly, the Digital Personal Data Protection (DPDP) Act, 2023 covers liability of crimes related to personal data, user content and data processing standards but does not prescribe any liability to AI systems which can invade data privacy by bypassing privacy measures. Thus, even with these modern legislations, India lacks a statute that directly addresses AI’s autonomous acts5. Currently there is no independent “AI Act” which addresses crimes specifically committed by the autonomous action of AI systems. This lacuna in law makes it challenging to establish guilt, prosecute, or deliver justice in cases involving autonomous AI systems. International Perspectives European Union (EU) The European Union has showcased the regulation of AI through it’s comprehensive EU AI Act. While the act doesn’t establish criminal liability to the autonomous actions of AI systems, 5 Sayyed, Hifajatali.

(2024). Artificial intelligence and criminal liability in India: exploring legal implications and challenges. Cogent Social Sciences. 10. 10. 1080/23311886. 2024. 2343195.

but it consists of preventative and civil measures, it classifies the various AI systems by risk levels, this ranges from minimal risk to high level risk, it makes compliance, safety checks and transparency mandatory, this ensures accountability. The EU also has an AI Liability Directive which complements the act, wherein presumption shifts part of the burden of proof to AI system operators. However this framework does not assign any criminal liability, only civil liability standards are applied. United States of America The United States of America as well does not have a separate federal statute which address criminal liability in crimes committed by autonomous AI systems. The criminal liability is decided on a case-to-case basis using product liability, tort law and existing cybercrime statutes.

The current framework does not account for actions done by AI autonomously that operate independently and without any human foresight. United Kingdom There is no concrete specific statute in the current legal system of the United Kindom (UK) as well which addresses the criminal liability in crimes committed by AI autonomously, independent from human intervention. The United Kingdom’s law commission has done extensive research on AI’s implications for criminal law. It’s reports highlight the difficulty of applying the elements of mens rea and actus reus in determination of criminal liability in cases of action by AI systems. The law commission has suggested in creation of hybrid accountability frameworks wherein both human developers as well as operators can be held partially liable depending on foreseeability and control.

UNESCO’s Role On the global scale UNESCO has issued the Recommendation on the Ethics of Artificial Intelligence. It consists on the need for emphasizing principles of fairness, transparency, human oversight and accountability. These guidelines are non-binding in nature, however they act as a guide as an ethical blueprint for member states, including India to align the development of AI with human rights and rule of law. This recommendation highlight the urgent need of legal reform for addressing AI-related harms, including potential criminal acts. India’s current legal framework alongside the global framework as well shows that criminal

liability in crimes involving AI systems still remain largely unaddressed in statutory law. While there exist certain civil remedies, but it doesn’t completely cover the responsibility gap, allowing autonomous AI systems to cause serious harm without any criminal liability to arise. This responsibility gap calls for an immediate legislative action to introduce new statutes or amendments to address the same. V.

THEORIES OF LIABILITY

To determine the liability for crimes committed by AI calls for re-examination of foundational legal doctrines. Due to lack of their physical form, consciousness, intention and other parameters of criminal liability according to traditional laws rooted in mens rea an actus reus becomes very difficult to apply. Legal theorists and policy makers explored various frameworks to attribute accountability. So theories like vicarious liability, product liability, corporate liability analogy, strict liability, and electronic personhood represent the principal approaches currently under debate. Vicarious liability This theory holds one person legally responsible for a wrongful act committed by another person which is mostly in employer- employee relationships.

The idea is that the person who is in a position of authority benefits from the agent’s actions and is best placed to bear the consequences of harm. So in context of AI, this theory can play a major role to assign liability to the user, operator or owner of an AI system for any wrongful act committed in the course of its operation. So if a company uses an AI chatbot that gives hate speech and misinformation, it can be argued that the company which means the employer could be held liable for the acts of its AI agent. The system operates under the company's direction to benefit it commercially, creating sufficient links to justify vicarious responsibility. This link can help us preserve the principle of accountability even when direct human intent is absent.

Obviously this model faces limitations because AI acts without morals unlike humans so this concept looks adaptable but is not a sufficient solution to the problem of accountability due to its unpredictable nature.

Product liability Under this theory manufacturers and developers are held responsible for the harm arising from defects in design, manufacturing or functioning of their products. Since AI is not legally recognised as a person if we look at it as a product rather than an agent, liability can be grounded in negligence or defective design principles. It particularly put emphasis on physical harms caused by autonomous systems like self-driving cars or robotic medical devices. If a self-driving car causes an accident due to an issue with the programming, the manufacturer could be liable for the production of a defective product which possesses unpredictable and unreasonable risk. This can be taken up in consumer protection law by the principle of ‘duty of care’ to ensure safety.

But obviously this theory has limitations in criminal contexts because product liability arises in civil law and not criminal law due to lack of mens rea. Moreover, AI systems keep on evolving through machine learning, so differentiating original design flaw and subsequent autonomous upgrades becomes extremely complex. So here the challenge is of identifying that at what stage the manufacturer's liability or responsibility ends. Corporate liability analogy This theory raises the question: can AI be treated similar to corporations which are considered as legal persons even after being artificial, can be prosecuted, fined or held liable because the law considers the mental and physical elements of crime to their human agents. Some argued that AI uses artificial intelligence to produce outcomes, sometimes without direct human input.

So if corporations can be considered legal persons despite they lack consciousness perhaps AI systems could also bear some responsibility. Corporations possesses assets and organizational structure capable of representing and regulating their actions, AI lacks all of it and therefore cannot be punished or rehabilitated which is the core purpose of criminal law. Therefore equating AI with corporate personhood remains legally unsatisfactory. Strict liability This doctrine removes the requirement of the concept of presence of mens rea so the liability arises solely from the commission of a prohibited act. It mostly focuses on the foreseeability

and preventability of harm. So under this theory developers or owners would be held responsible for offences cause by this machine, irrespective of fault. This ensures responsibility which would encourage the liable parties to high standards of care, security and oversight. So if an AI driven drone unintentionally crosses airspace regulations leading to damage, the operator could be held liable even without negligence. This doctrine mostly promotes deterrence by precautionary behavior which is extremely necessary in this full AI driven era which could lead to unpredictable damages. But excess emphasis on strict liability can hinder innovation by exposing developers to unnecessary legal risks. So strict liability should apply to high risk AI systems but not ordinary applications for a balanced solution.

Each theory gives us a direction to partial solutions but none resolves the confusion. Some theories ensure human accountability and some talks about the extension of legal identity. So a composite approach can help uphold justice and deterrence in a world where AI is ruling and the line between this system and human continues to blur. VI.

ETHICAL & PHILOSOPHICAL DIMENSIONS

Rapid dependency on AI as an autonomous decision maker compels a fundamental reexamination of the moral foundations of criminal liability. The foundation of criminal jurisprudence lies in the concept of moral agency which refers to the capacity to differentiate wrong from right and to bear responsibilities for your action. Unlike humans, AI remains devoid of consciousness or moral understanding of things. They function solely based on algorithmic logic. So they have no intention to cause harm. They merely carry out operations which takes us to another important question: can an entity incapable of moral judgement be subject to blame? The theory of punishment is necessary and works only during the presence of guilt, remorse and the capacity for deterrence, these qualities can’t be possessed by AI.

so holding a machine accountable will only displace the ethical burden from human to machine, leaving victims remediless. Therefore, liability rules should mostly focus on preventing harm rather than punishing AI itself. So the goal becomes risk minimization to forestall future harm. Every automated system is the product of human choices. From data selection to deployment, it's all a human brain so it makes sense if they are held accountable emphasizing on their duty of care. These gaps between the liability of a human and AI will keep on increasing if there is

no such action taken to ensure proper justice to the victims. In conclusion the future of AI must rest on transparency and moral responsibility and not machines, but of the humans designing and depending on it. VII.

RECOMMENDATIONS & REFORMS

With the increase in usage and developments of AI, it demands a concrete framework that closes the existing responsibility gap between human oversight and machine action. The current legal framework in India is insufficient to address crimes committed by autonomous AI systems. Therefore, it is imperative that a separate statute must be introduced which defines terms such as accountability, ethical governance, oversight mechanisms and addresses criminal liability in crimes committed by AI systems. There must be mandatory human supervision for autonomous AI systems and high risk AI systems. AI systems particularly used in finance, medical, military, law enforcement must be operated strictly under human supervision. This prevents the risk of unsupervised behaviour by AI and ensures humans have the ultimate decision-making power.

All autonomous AI systems must have mandatory registration under a designated authority ensuring compliance of ethical and moral standards and provide for liability insurance as well. This will ensure victims receive compensation for the harm suffered. A National AI Audit must be established for overseeing behaviour of AI systems as well as

compliance of AI systems. This authority will be responsible for conducting regular algorithmic audits, ensuring transparency, data protection, and bias detection across all deployed AI systems. It will maintain a registry as well of the high risk AI technologies and enforce penalties for non-compliance. This measure will result in increased accountability and will act as a preventive mechanism as well. The proposed reform measures will collectively ensure in creation of a balanced framework that promotes innovation along with accountability and a just legal regime which upholds human oversight and responsibility in today’s highly autonomous world.

VIII. CONCLUSION

The rapid evolution of artificial intelligence challenges the basic foundation of criminal

jurisprudence. The current Indian law mainly focuses on human intention and moral liability, which results in confusion to accommodate the self dependent and emotionless nature of AI systems. AI cannot comprehend mens rea or the moral dimensions of its actions leaving a huge gap by fully declining the traditional concept of criminal liability but absolutely causing real and tangible injury. Responsibility therefore should be redirected towards those who manufactured, trained and designed them. It's the humans who profit from their operation so it should be them who should be held accountable in case of an accident. Since AI is playing a major role in the economy, laws must evolve to balance innovation with accountability with complete fairness.

Hence, the courts must ensure that in a world where minds are made of code, justice must remain distinctly human and accountability remains firmly in the human hands. Mandatory human supervision should be enforced for all high risk crimes with the use of AI specially in the sensitive domains such as finance, medicine, military, and law enforcement to ensure human control over critical decisions. Humans should not solely rely on AI and to keep track of AI’s working national AI audit authority should be established for continuous oversight, algorithmic audits, bias detection and transparency enforcement. Mandatory registration and liability insurance in the context of automated AI systems would guarantee compensation to victims of AI- related crimes. Therefore these reforms would balance the ecosystem which believes in promoting innovation as well as ensuring accountability and justice.