Document Type : Research Paper

Author

• Heidar Piri

Assistant Professor of Public International Law, Law Department, Faculty of Humanities and Social Sciences, University of Kurdistan, Sanandaj, Iran.

Abstract

Introduction
The rapid development and proliferation of Artificial Intelligence (AI) systems, particularly in military and security domains, represent a paradigm shift with profound implications for international law. While offering potential benefits in efficiency and capability, the autonomous nature of advanced AI systems raises acute legal and ethical challenges, especially concerning accountability for serious violations of international law. The deployment of AI in armed conflict, such as through Lethal Autonomous Weapon Systems (LAWS), and its use in contexts that may facilitate international crimes, like genocide, war crimes, and crimes against humanity, necessitates a fundamental re-examination of traditional legal frameworks. Current international criminal law is fundamentally anthropocentric, built upon the principle of individual criminal responsibility which requires both a physical act (actus reus) and a guilty mind (mens rea). This framework struggles to accommodate entities that can act independently, learn from their environments, and cause significant harm without direct, predictable human intervention at the moment of the act. This article delves into the core legal dilemma: who can and should be held accountable when an AI system commits an act that constitutes an international crime? It explores the feasibility and necessity of attributing criminal liability directly to AI systems themselves, potentially by granting them a novel legal status such as electronic personality (e-person), as opposed to, or in conjunction with, holding the various human actors in their chain of development and deployment responsible.
 
Research Question(s)
This research is guided by the following primary questions:

In the context of international crimes, who can be held accountable for proscribed acts or crimes committed by AI systems? Are there any grounds for labeling them as criminals or granting the status of e-person?
What are the legal and philosophical grounds for, and obstacles against, granting autonomous AI systems a form of legal personality (e-personhood) to bear rights and obligations, including criminal liability?
Can the existing requirements for establishing criminal responsibility under international law, particularly the actus reus and mens rea, be satisfied by AI systems in their current or foreseeable state of development?
What alternative models of liability, such as holding programmers, manufacturers, military commanders, or states responsible, are available and effective under current international law, and what are their limitations?
What are the potential solutions and necessary legal reforms, including at the level of the ICC Statute, to address the accountability gap posed by AI systems capable of committing international crimes?

 
Methodology
The article has been performed based on the descriptive and analytical research method. The necessary data has been collected by library method. The research adopts a critical and forward-looking approach, analyzing the coherence and sufficiency of existing legal doctrines, identifying conceptual gaps, and proposing normative solutions based on logical reasoning, comparative analysis of analogous legal constructs (e.g., corporate criminal liability), and the functional demands of international justice.
 
Results
The investigation yields several key findings:

The Case for Electronic Personality and Direct AI Liability:Arguments for granting AI systems a form of legal personality is compelling, drawing parallels with the historical extension of legal personhood to corporations. Proponents argue that highly autonomous AI, capable of independent analysis, decision-making, and learning, possesses a functional equivalence to the rational agency required for responsibility. The concept of "electronic responsibility" is presented as a necessary tool to prevent human actors from evading liability by hiding behind the complexity and autonomy of machines.
Significant Legal Obstacles:The path to direct AI criminal liability is fraught with major hurdles under current law:

Anthropocentric Foundations:The Rome Statute and the general principles of international criminal law are firmly rooted in human agency. Terms like "person" are interpreted as natural persons.
The Mens ReaRequirement: The most formidable barrier is the mental element. While an AI system can arguably satisfy the actus reus (the physical act), attributing intent, knowledge, or recklessness—subjective mental states tied to consciousness, moral understanding, and foresight—to a machine remains deeply problematic both legally and philosophically.
Lack of Legal Personality:AI systems currently lack recognized legal personality in international law, a prerequisite for being a subject of rights and duties, including criminal liability.
Punishment Incommensurability:Traditional penal theories (retribution, deterrence, rehabilitation) lose meaning when applied to non-human entities that cannot feel guilt, suffer, or be morally reformed.


Analysis of Alternative Human Liability:In the absence of direct AI liability, the focus shifts to human actors. However, attributing responsibility to programmers, manufacturers, operators, or military commanders is often hampered by practical and legal difficulties: the problem of many hands, the challenge of proving individual mens rea for unforeseeable autonomous actions, and the potential lack of "effective control" required for command responsibility when dealing with learning systems.

The article concludes that a multi-pronged approach is needed:

Strict Liability Models:The ICC and international criminal law may need to embrace forms of strict or no-fault liability for situations involving autonomous systems, moving away from mens reaas an absolute central pillar for certain contexts.
Regulation and Prohibition:Strengthening IHL compliance through rigorous legal reviews of new weapons (Article 36, AP I), enhancing precautionary measures, and potentially negotiating treaties to limit or ban certain types of autonomous weapons.
Statutory Reform:For direct AI liability to become viable, the Rome Statute would require amendment. Articles 1 and 25(1) could be revised to explicitly extend the Court's personal jurisdiction to legal persons or electronic persons, and a new framework for electronic responsibility would need to be codified.
Ethical and Technical Safeguards:Implementing robust ethical guidelines for developers, incorporating IHL rules directly into AI training (law encoding), and creating reliable fail-safe mechanisms for deactivation.

 
Conclusion
The advent of AI systems with significant autonomy presents one of the most profound challenges to the international criminal justice system. While the theoretical appeal of holding AI directly accountable, grounded in its capacity for autonomy, independent analysis, decision-making, and a functional approximation of intentionality, is as compelling as arguments for extending legal personality to other non-human entities, it is currently precluded by foundational legal principles. The core impediments are the irreconcilable absence of a mens rea in AI and the lack of an established legal personality, making the imposition of direct criminal responsibility neither theoretically coherent nor practically feasible under the extant anthropocentric framework. Consequently, in the near term, the most viable accountability mechanisms must focus on reinforcing the responsibility of the myriad natural and juridical persons involved in the construction, programming, training, and deployment of AI systems for their malfunction or unlawful outcomes. However, this human-centric approach is itself severely constrained by traditional mens rea requirements, creating a significant liability gap and risking an anomaly where international criminal law becomes ineffective in addressing harms caused by this technology, thereby undermining human rights protection and international justice.
To bridge this gap, a dual-path strategy is essential. Immediately, it necessitates strengthening preventive International Humanitarian Law (IHL) regulations and rigorously applying existing models of human responsibility across the AI lifecycle. Simultaneously, for the future, the international legal community must engage in proactive and principled reform. As AI evolves towards greater sophistication, the pressure to reconceptualize legal personhood will intensify. The ICC will only be able to effectively prosecute crimes involving autonomous AI agents if it embraces legal innovations such as strict liability and other alternative models of fault-based liability, which have so far been marginalized. Ultimately, ensuring accountability and preventing impunity for the gravest crimes may require international criminal law to transcend its strict anthropocentrism and incorporate a functional, graduated model of electronic responsibility. This represents a seismic shift in legal philosophy, demanding careful preparation and reasoned debate to develop the unique formula necessary for a future where the law keeps pace with technological agency.

Keywords

• Electronic responsibility
• AI
• Autonomous weapons
• International Criminal Court
• Legal personality
• International crimes

Main Subjects

• International Criminal