Criminological Risks and Legal Aspects: When AI Crosses the Line

Executive Summary

As AI transitions from simple tools to autonomous decision-makers, the boundary between "malfunction" and "crime" is blurring. This paper, authored by researchers from the Kazan Innovative University, tackles one of the most uncomfortable questions in modern law: Who is responsible when a self-learning machine commits a crime?

The authors position this work as a critical theological and legislative intervention, moving beyond technical optimization to address the "reverse side" of the AI coin. They argue that the legal world is fundamentally unprepared for the technological singularity—the moment AI intelligence surpasses human capacity.

The "Mens Rea" Dilemma: Why Current Laws Fail

Traditional criminal law is built on human psychology. To convict someone, you usually need to prove two things:

1. Actus Reus: The physical act of the crime.
2. Mens Rea: The "guilty mind" or criminal intent.

How do you prove "intent" in a neural network? If a self-driving car decides to protect its passenger at the expense of a pedestrian (the "Moral Machine" problem), is it a calculation error or a deliberate choice? The paper highlights that current anthropocentric laws create a liability vacuum where neither the developer nor the user can be held fully responsible for the emergent behaviors of a self-learning system.

Methodology: Mapping Criminological Risks

The researchers categorize the threats posed by AI into a structured framework:

1. Direct Criminological Risks

• Autonomous Decisions: AI independently chooses to perform an action that qualifies as a crime (e.g., identity theft or unauthorized financial transfers).
• Criminal Hijacking: Intentional tampering with AI software to cause social harm.
• AI as a Weapon: Systems specifically designed by criminals to automate cyber-attacks.

2. Indirect Criminological Risks

• Developer Negligence: Coding errors that manifest only in specific, high-stakes environments.
• Operational Errors: Unintended "hallucinations" or logical failures during execution, such as the reported errors in medical AI like IBM Watson Health.

Note: The integration of AI involves a complex interplay between technology, social ethics, and legal regulation.

The Solution: Legal Fiction and the "Electronic Person"

The paper’s most provocative proposal is the use of Legal Fiction.

Instead of trying to prove a machine has a "soul" or "consciousness," we should treat it as a Legal Entity, similar to a corporation. A corporation can be sued and penalized even though it isn't a breathing person. The authors suggest:

• Limited Legal Personality: Giving AI rights and duties separate from its creators.
• Technical Sanctions: Instead of prison, "punishment" for AI would involve deactivation, reprogramming (a digital lobotomy), or conferring "Criminal Status" to warn other users.

Experimental Insight: The Moral Choice

The paper references the MIT "Moral Machine" study (2.3 million responses), proving that human ethics are not universal—they are influenced by religion, gender, and culture. If we cannot agree on a human ethical framework, hard-coding "ethics" into AI becomes an impossible task. This confirms the authors' view that AI must be recognized as a "source of increased danger," requiring mandatory insurance funds and registration systems.

Critical Analysis & Future Outlook

While the "legal fiction" model solves the immediate problem of financial liability, it still struggles with the concept of retributive justice. Can we ever truly "punish" an algorithm?

Key Takeaways:

• The 15-Year Horizon: Within the next decade, a total revision of intellectual property and criminal law is inevitable.
• Systemic Safety: The paper calls for a "safety switch" or "immediate shutdown" protocol for all autonomous systems in case of force majeure.

Conclusion: As AI continues to evolve, we must shift our focus from "Can we build it?" to "Can we govern it?" Resolving the legal personality of AI is not just a theoretical exercise for lawyers—it is a survival requirement for society.

Future legal frameworks must balance innovation with a rigorous ethical and criminological "kill-switch" protocol.