نوع مقاله : مقاله پژوهشی

نویسندگان

• نورا احسانگر ¹
• علیرضا یزدانیان ²

¹ استادیار گروه حقوق خصوصی، دانشکدۀ ادبیات و علوم انسانی، دانشگاه شهرکرد، شهرکرد، ایران.

² دانشیار گروه حقوق خصوصی دانشکده علوم اداری و اقتصاد دانشگاه اصفهان، اصفهان، ایران

چکیده

این پژوهش به تحلیل مبانی حقوقی مسئولیت مدنی ناشی از عملکرد سامانه‌های هوش مصنوعی، با تمرکز بر قواعد عمومی مسئولیت مدنی در حقوق ایران و فرانسه می‌پردازد. در آغاز، با تحلیل ویژگی‌های ماهوی هوش مصنوعی، این فناوری به‌عنوان شیئی دارای نقش فعال و اثرگذار در ایجاد خسارت تلقی شد. سپس، دو معیار برای انتساب مسئولیت مورد بررسی قرار گرفت: مسئولیت مبتنی‌بر مالکیت و مسئولیت مبتنی‌بر نگهداری. در فرض وحدت مالک و کاربر، تفاوتی میان اعمال دو معیار وجود ندارد و مسئولیت به شخص واحد قابل انتساب خواهد بود. اما در فرض تفکیک، معیار «نگهداری» و سلطۀ عملی در مواردی که کاربر در زمان وقوع خسارت نقش فعال دارد، اقناع‌کننده‌تر به‌نظر می‌رسد. درمقابل، اگر خسارت ناشی از فعل غیرقابل پیش‌بینی و مستقل هوش مصنوعی باشد، نظریۀ مسئولیت محض و مبتنی‌بر مالکیت، توجیه حقوقی قوی‌تری خواهد داشت. از این منظر، مسئولیت مدنی در حوزۀ هوش مصنوعی، ماهیتی ترکیبی می‌یابد و بسته به شرایط، می‌تواند بر یکی از دو مبنای یادشده استقرار یابد. یافته‌ها نشان می‌دهد تا پیش از تصویب قوانین خاص، باید رویکردی انعطاف‌پذیر و متناسب با نوع رابطه و شرایط حادثه اتخاذ گردد تا ضمن جلوگیری از نتایج ناعادلانه، عدالت جبرانی محقق شود و حقوق مسئولیت مدنی با واقعیت‌های فناوری نوین هماهنگ گردد.

کلیدواژه‌ها

• کاربر
• مالکیت
• مسئولیت مدنی
• نگهداری
• هوش مصنوعی

موضوعات

• حقوق خصوصی

عنوان مقاله [English]

On Ownership or Custody: A Comparative Study in French Law

نویسندگان [English]

• noura ehsangar ¹
• Alireza Yazdanian ²

¹ Assistant Professor, Department of Law, Faculty of Literature and Humanities, Shahrekord University, Shahrekord, Iran

² Associate Professor of Private Law, Faculty of Administrative Sciences and Economics, University of Isfahan, Isfahan, Iran

چکیده [English]

Artificial intelligence (AI), as one of the foundational technologies of the modern era, has profoundly affected various dimensions of individual and social life. Features such as autonomous decision-making, continuous learning capabilities and the unpredictability of outcomes have posed new challenges for legal systems. Among these, civil liability arising from the operation of AI in non-contractual relations has emerged as one of the most pressing legal questions. In particular, within the Iranian legal system, where no specific regulations on this subject currently exist, the determination of liability requires a reinterpretation of general principles and traditional rules of responsibility. The central question is whether civil liability should be attributed on the basis of ownership of AI systems or, rather, grounded in standards of maintenance, care, and proper use.
This study, with a primary focus on Iranian law and a comparative reference to French law, seeks to clarify the legal nature of AI and to identify the basis of civil liability arising from its operation. The inquiry begins with the question of AI’s legal status: should it be regarded as akin to a human endowed with reason and will, or should it be treated as an object or, in some cases, analogized to an animal? It then examines two main criteria of liability: ownership-based liability and liability grounded in maintenance and care. Finally, the legal relationship between owner and user is analyzed, and the findings are synthesized through different hypothetical scenarios.
The methodology employed is descriptive–analytical, relying on documentary research. Data are drawn from Iranian statutes, legal doctrine, and French law. The focus is on general rules of civil liability contained in the Iranian Civil Code, the Law of Civil Liability, and the Islamic Penal Code, which regulate responsibility for objects, animals, and buildings, and which can be extended to AI systems.
Since considering AI as a human requires attributes such as will, conscience and rationality, features that AI does not yet possess, the more realistic approach is to treat AI as an “object” capable of ownership and control. Accordingly, the framework of liability for objects provides a suitable analytical lens. Within this framework, two approaches emerge:

Ownership-based liability: Under the strict liability theory, mere ownership of an object (here, the AI system) suffices to establish responsibility, even if the owner had no involvement in the harm caused. This approach is justified by the aim of maximizing protection for victims.
Maintenance and proper use-based liability: Here, liability depends on the fault or negligence of the user or possessor. The key element is actual control and the duty of care at the time the damage occurs. This approach aligns with Article 334 of the Iranian Civil Code and Articles 512, 522, and 524 of the Islamic Penal Code, which emphasize the roles of “possessor” and “controller.”

The central issue is whether liability requires the unification of ownership and use in a single person, or whether these attributes, when separated, shift responsibility to different parties. Two situations are examined:

Unity of owner and user: In this case, no distinction arises between the two criteria, since both legal and practical authority are vested in one person. Iranian law, particularly Article 333 of the Civil Code, establishes that mere ownership coupled with effective control suffices for liability. Analogy with liability for animals reinforces this conclusion.
Separation of owner and user: Two scenarios emerge:


Damage due to the user’s fault: If the user exercises actual control and has a duty of care at the time of the accident, liability is attributed to the user. This position is consistent with the theory of fault in maintenance and with French jurisprudence.
Damage due to unforeseeable and autonomous behavior of AI: Where harm results from AI’s self-learning or self-correcting behavior, beyond the foresight of either owner or user, the fault-based model loses its explanatory power. Two competing views arise:


Responsibility of the owner under strict liability: Pursuant to Article 523 of the Islamic Penal Code, even when use is authorized, the owner remains liable unless the causal link is severed. This approach simplifies the identification of the liable party but risks unfair outcomes.
Transfer of responsibility to the user as the actual possessor: Drawing on Articles 334 of the Civil Code and 512 and 522 of the Islamic Penal Code, which highlight “possessor” and “controller,” this view argues that liability should fall on the party exercising real control. This prevents imposing liability on an owner who is effectively disconnected from the harmful act and aligns more closely with principles of fairness.

The contribution of this study lies in proposing a hybrid model that allocates civil liability for AI according to the circumstances of the harm: liability falls on the user in cases of negligent maintenance or misuse, while in cases of unforeseeable and autonomous AI behavior, strict liability of the owner or a functional transfer of liability to the user may apply.
Ultimately, civil liability in the field of AI cannot rest solely on either ownership or maintenance; rather, it requires a nuanced combination of both. Iranian and French law, through existing provisions, provide avenues to extend traditional rules to these new contexts. However, to avoid unjust outcomes, legislators must move beyond traditional frameworks and adopt specific regulations for AI, including mandatory insurance schemes, explicit duties for users, and protective mechanisms for victims.

کلیدواژه‌ها [English]

• Artificial Intelligence
• Civil Liability
• Ownership
• Custody
• User