Criminal Liability for AI-Generated Content
1. Introduction: The Impact of the Digital Revolution on Criminal Law
The development of artificial intelligence (AI) technologies is progressing at a pace not foreseen by legal systems. Text, images, sound, video, and even full-fledged news or scientific articles can now be produced entirely by algorithms. This situation seriously challenges the traditional concepts of perpetrator, act, intent, and consequence in Turkish criminal law. In the Turkish Penal Code No. 5237, the perpetrator is defined as the natural person who performs the act or who carries out the act under their control. However, artificial intelligence systems do not possess "will" or "culpability" on their own. Therefore, AI itself does not have criminal responsibility. Nevertheless, if content produced by an AI system constitutes a crime, responsibility will be sought from the humans who use, manage, operate, or design that system.
Criminal law is essentially a system of liability based on fault. The inability to consider artificial intelligence as a criminal offender necessitates the identification of human intent behind the act. In this context, the question of "who is the perpetrator?" constitutes the most complex point in practice. This is because human intervention in the production of AI content can sometimes be overt, sometimes indirect, and sometimes completely invisible. In cases where the identity of the perpetrator cannot be determined, criminal liability does not arise in principle; because, according to the principle of "personal responsibility," no one can be held responsible for the actions of another (Turkish Penal Code, Article 20/1).
2. Legal Nature of AI-Generated Content
How to legally classify content produced by artificial intelligence is of fundamental importance in determining criminal liability. According to Article 1/B of the Law on Intellectual and Artistic Works (FSEK), a work is an intellectual and artistic product bearing the distinctive characteristics of its owner. However, artificial intelligence is not a person, and the concept of "distinction" refers to the creative personality of a human being. Therefore, a fully automated AI-generated text or image cannot be considered a "work" in the classical sense. However, such content qualifies as "content" under Law No. 5651. Consequently, any text, video, or image uploaded to the internet or presented to a user by artificial intelligence is considered "internet content," and in this case, the title of "content provider" is assigned to the person managing or using it.
Article 2 of Law No. 5651 defines a "content provider" as the person who produces, modifies, or provides any information or data made available on the internet. According to this provision, the person or institution operating an AI system, not the AI itself, should be held responsible for content generated by that system. For example, a person who publishes text generated by ChatGPT or a similar system on their website is legally considered the "provider" of that content. Therefore, a human actor is always at the beginning of the chain of criminal responsibility.
3. Artificial Intelligence from the Perspective of the Fundamental Principles of Criminal Law
From the perspective of the systematic approach of the Turkish Penal Code, in cases involving crimes stemming from artificial intelligence , the principles of culpability , personal responsibility , legality , and fault must be examined first. According to the principle of culpability (Articles 21-22 of the Turkish Penal Code), for the perpetrator to be punished, they must have committed the act at least negligently. Since artificial intelligence does not possess free will, it cannot be subject to the concepts of intent or negligence; however, whether the person who programmed or used it is culpable in this context is evaluated separately in each specific case. Intent is when the perpetrator knowingly and willingly commits the crime. For example, if a user instructs artificial intelligence to "produce a news article containing defamation about so-and-so," the element of intent is clear. However, if the content is erroneous due to incorrect learning by the artificial intelligence, then the provisions of negligence or negligent liability come into play.
The provisions on complicity in Articles 37-39 of the Turkish Penal Code are also important in this area. If more than one person has contributed to the artificial intelligence system producing criminal content, criminal responsibility is shared as direct perpetrators, instigators, or accomplices. At this point, corporate security measures (Article 60 of the Turkish Penal Code) can be applied to the companies operating the artificial intelligence platform. Because even though legal entities are not considered criminal perpetrators, they can be subject to sanctions such as confiscation or revocation of operating licenses.
4. Major Crime Types Associated with AI Content
4.1. Insult and Defamation (Turkish Penal Code Articles 125 and 267)
If AI-generated automated news or social media posts contain defamatory or untrue statements targeting individuals, the mental element of the crime is considered to have been fulfilled by a human perpetrator. The Supreme Court of Turkey states that the perpetrator's intent in social media posts should be evaluated in conjunction with their will to verify and control the post. This jurisprudential logic also applies to AI-generated content. If a user shares the generated content without verifying it, they may be held liable under the category of "conscious negligence.".
4.2. Unlawful Processing of Personal Data (Turkish Penal Code Articles 135–140, Personal Data Protection Law Article 12)
If personal data is processed without consent or not anonymized in datasets used for AI training, the data controller may face criminal liability. Especially in cases involving the use of facial features, voice, or identity information, data controllers who fail to take the necessary technical and administrative measures in accordance with Article 12 of the Personal Data Protection Law may face both administrative fines and imprisonment under Article 136 of the Turkish Penal Code.
4.3. Obscenity and Child Abuse (Turkish Penal Code Article 226)
If AI models that generate visual content can create obscene images involving children, the perpetrator is directly the person who produces and disseminates this content. Article 226/3 of the Turkish Penal Code explicitly criminalizes the production and sharing of obscene images of children. In this case, AI is a tool; the person using it or enabling the creation of this output is the perpetrator.
4.4. Forgery and Fraud (Turkish Penal Code Articles 204–207, 158)
Producing fake identities, diplomas, licenses, or documents using AI constitutes "document forgery," while obtaining benefits through voice or image imitation constitutes "fraud through information technology." At this point, it is clear that AI is not the perpetrator, but rather the tool used in committing the crime.
4.5. Publicly Disseminating Misleading Information to the Public (Turkish Penal Code Article 217/A)
If the false news or manipulative information produced by AI has the potential to disrupt public peace, the perpetrator may face liability under both Article 217/A of the Turkish Penal Code and Articles 5-6 of Law No. 5651. The platform administrator's failure to prevent the dissemination of such content, which clearly violates the law, constitutes secondary fault.
5. Identifying the Perpetrator and the Chain of Responsibility
In artificial intelligence systems, the perpetrator of a crime can be a multiple-pronged individual. Firstly, the useris the primary perpetrator if they directly generated content through commands given to the AI and knew that this content constituted a crime. The developeris negligently liable if they failed to take sufficient precautions to prevent the system from generating criminal content. The platform administrator is considered secondarily liable if they failed to implement a warning-and-removal mechanism.
In practice, the criterion of "predictability" is used to determine the boundaries of all three roles. That is, it is examined whether the occurrence of harm or a crime was reasonably foreseeable. If the developer knew of the system's potential to produce hate speech, obscenity, or disinformation and yet failed to install the necessary filters, their fault is aggravated. Conversely, if criminal content arises as a result of a completely unforeseeable technical malfunction, no criminal liability arises.
6. Issues of Evidence, Proof, and Digital Examination
The principle of freedom of evidence applies in criminal proceedings; however, the reliability and integrity of digital evidence are of paramount importance. In AI-generated content, the chain of evidence is established through "log records," "prompt history," "model version," "IP addresses," and "timestamps." According to Article 134 of the Code of Criminal Procedure, data obtained from information systems can be subjected to forensic digital examination. The crucial point here is to maintain the chain of evidence and verify the hash values. Furthermore, the provisions of the Personal Data Protection Law regarding the protection of personal data must also be observed during the evidence gathering phase.
During a trial, it is impossible to determine fault without identifying who directed the AI-generated content, which commands were used to generate it, and how the sharing chain developed. Therefore, expert examinations require collaboration between both software experts and legal professionals. Otherwise, the technical complexity of the legal process may hinder the uncovering of the material truth.
7. International Approaches and Comparative Perspectives
The European Union, its Artificial Intelligence Act (AI Act) , imposes clear obligations on developers and distributors of high-risk systems. While this regulation does not directly foresee criminal sanctions, it will serve as a guide in Turkish law in terms of "predictability" and "duty of care." In US law, criminal liability for artificial intelligence is denied under the "mens rea" concept; however, the developer can be sued under the "negligent supervision" doctrine. In the UK, under the "intermediary liability" regime, AI-based platforms are subject to a limited liability regime as "intermediary service providers." The orientation of Turkish law is largely along these lines: the principle of supervision and duty of care instead of direct criminal liability.
8. De Lege Ferenda Recommendations: Need for Legislation Update
As artificial intelligence technologies continue to develop, existing legislation is proving insufficient. The Turkish Penal Code needs to be amended to include an article defining "crimes committed through automated systems" and clearly regulating the criteria for culpability. It is also essential to update Law No. 5651 with specific notice-and-takedown procedures for AI-based content creators and to expand the algorithmic monitoring obligations of social network providers. Furthermore, clarifying the distinction between data crimes and administrative violations within the Turkish Penal Code and the Personal Data Protection Law (KVKK) is crucial for defining the limits of predictability. Universities and bar associations should create expert witness lists in the field of "AI and Criminal Law," and the technical support units of prosecutor's offices should be strengthened.
9. Conclusion: The New Balance of Criminal Justice in the Age of Artificial Intelligence
Artificial intelligence is fundamentally transforming the criminal law system. Currently, it is not possible to directly punish artificial intelligence as an offender; however, liability may arise within the limits of human intent, supervisory duties, and predictability. In this process, a delicate balance must be struck between the rule of law, freedom of expression, and personal rights. The aim of criminal law is not to suppress innovation, but to define boundaries that protect public safety.
Therefore, in the age of artificial intelligence, the concept of "criminal responsibility" now encompasses not only human actions but also digitally driven actions. The fundamental task facing Turkish law is to create a normative framework appropriate to this new reality and to ensure the relevance of justice in the face of technological transformation.
About Post Author
