1. Abstract
This article
examines acts committed through artificial intelligence (AI) from the
perspective of Turkish Penal Code. Since AI is not recognized as a direct
perpetrator, criminal liability is primarily attributed to human actors such as
developers, users, and system owners. In light of international examples and
recent developments, the legal challenges encountered in AI-generated
content—particularly with systems like Grok—are analyzed. The study reveals
that current Turkish Penal Code regulations do not adequately cover AI-specific
acts, leading to significant normative gaps. Emphasis is placed on the
necessity of enacting effective legal regulations in this field and aligning
national standards with international norms.
Keywords : Artificial
Intelligence, AI and Penal Code, Criminal Liability, Concept of Perpetrator,
Intent and Negligence, Principle of Legality, AI-Generated Crimes, Legal
Responsibility, Algorithmic Crimes, Grok AI, AI Content Generation.
2. Introduction
Artificial
intelligence (AI), one of the most striking technological advancements of the
21st century, has permeated nearly every aspect of human life and led to revolutionary
transformations across numerous sectors. AI systems, which are actively
employed in fields such as healthcare, transportation, education, and finance,
have evolved from mere auxiliary tools into autonomous actors that actively
participate in decision-making processes. These technological developments not
only impact the socio-economic structure but also directly affect legal
systems.
Penal Code is
fundamentally built upon a system of liability that is based on human will and
culpability. Within this framework, the perpetrator of a crime is traditionally
a natural person who acts with intent or negligence. However, the growing
capacity of AI to make autonomous decisions necessitates a reassessment of
classical concepts in penal code theory, such as "perpetrator,"
"fault," "intent," and "negligence." The legal
classification of an act committed by an AI system, the identification of the responsible
party, and the determination of criminal liability remain questions without
clear and definitive answers under the current legal paradigm.
The assessment of
acts committed through AI within the context of penal code entails significant
normative and practical gaps and controversies. This article aims to explore
how AI should be approached within the framework of penal code; to what extent
and to whom criminal liability may be attributed in acts committed via such
technologies; how the existing legal system responds to these acts; and in what
ways it falls short.
3. The Concept of
Artificial Intelligence and Its Legal Characterization
Artificial
intelligence (AI) can generally be defined as systems that emulate human
intelligence. These systems are capable of performing cognitive processes such
as learning, reasoning, problem-solving, perception, and even language
comprehension through algorithms. From a technical perspective, AI is often
evaluated within the subfields of machine learning, deep learning, neural
networks, and natural language processing. However, the primary concern lies, for
the legal discipline, in the decision-making capabilities of such technologies
and how these decisions are to be evaluated within the legal order.
AI systems may be
classified according to their level of technological advancement as assistive
systems, semi-autonomous systems, and fully autonomous systems. While assistive
systems function with human intervention, fully autonomous systems possess the
capacity to make and execute decisions independently of human control. It is
primarily this latter category that gives rise to legal challenges. The
fundamental issue is determining who bears responsibility for the decisions or
acts executed by these systems—a matter that cannot be directly answered by
traditional legal theories.
Under Turkish law,
there is no explicit regulation concerning the legal status of artificial
intelligence. According to the civil law system in Türkiye, only natural and
legal persons possess legal capacity. In this context, AI is not regarded as a
legal subject but rather as a technical tool or object. However, this approach
has become insufficient with the increasing complexity and autonomy of AI systems.
In particular, legal systems built upon the notions of culpability and
intent—such as penal code—face normative gaps as a result.
One of the most
fundamental issues from a penal code standpoint is whether an act committed by
AI can be classified as a legal “actus reus”. Under penal code, an act
is generally defined as a voluntary human action. This brings forth numerous
technical and ethical questions, such as whether AI can make independent
decisions, to what extent such decisions are foreseeable, and how much they can
be influenced or controlled. Therefore, the legal characterization of AI
requires a multidisciplinary approach that includes not only legal scholars but
also ethicists, computer scientists, and sociologists.
Penal code is a
discipline grounded in the principle of liability based on human intent. For an
act to constitute a crime, it must be explicitly defined in law, the
perpetrator must have acted with intent or at least negligence, and the act
must be unlawful. However, the rise of AI technologies significantly challenges
these core principles of penal code.
4. Artificial
Intelligence in the Context of Fundamental Principles of Penal Code
4.1. The Principle
of Legality and the Problem of Legal Certainty
One of the cardinal
principles under penal code is the principle of legality ("nullum
crimen, nulla poena sine lege"). According to this principle, an act
can only be deemed criminal if it has been clearly defined as such by law prior
to its commission. However, AI systems are capable of making decisions that are
unforeseeable and may give rise to novel forms of conduct. For instance, in the
case of an autonomous vehicle striking a pedestrian without a driver, the
applicable legal provision and the liable party are not clearly defined in the
law. This situation contradicts the principles of legal certainty and foreseeability
that are essential to criminal justice.
4.2. The Concept of
the Perpetrator and Capacity for Fault
Under penal code,
the perpetrator is the individual who commits the act that legally constitutes
the crime. This person must be an entity that acts with volition and intent and
possesses legal capacity. However, AI
lacks legal personality and any capacity for fault. In other words, AI cannot
be held liable as a perpetrator. At this point, questions arise as to whether
liability should instead fall upon the programmer, the user, or the legal
entity that owns the AI system.
4.3. Evaluation in
Terms of Intent and Negligence
Intent refers to
the perpetrator knowingly and willingly committing the act. Negligence, by
contrast, involves any failure to exercise due care. AI systems are incapable
of possessing either element, as they cannot form conscious decisions nor
breach duty of care. Nevertheless, if an AI system is known to function
erroneously in a foreseeable manner, users may be held liable for negligence if
they continue to employ it. For example, continuing to use a defective AI
system despite knowledge of its flaws may have significant legal consequences.
4.4. The Causal
Link and the Problem of Foreseeability
Many AI systems
operate as "learning systems" that evolve over time and adapt their
decision-making processes. This feature makes it increasingly difficult to
establish a causal link between an act and its outcome. From a penal code perspective,
causality requires a direct connection between the perpetrator and the result
of the crime. However, when an AI system autonomously learns and makes an
erroneous decision, it remains unclear who should bear responsibility and to
what extent.
5. Criminal Liability
in Acts Committed Through Artificial Intelligence
The complex and
autonomous nature of AI systems raises significant questions regarding who
bears responsibility when unlawful acts are committed through such systems.
According to the principle of individual responsibility under penal code,
liability arises for the person who personally commits an act that fulfills the
elements of a crime. However, since AI cannot be regarded as a direct
perpetrator, determining responsibility often requires a multifaceted analysis
involving multiple actors.
5.1. Can Artificial
Intelligence Be Considered a Perpetrator?
Due to the lack of
legal personality, artificial intelligence cannot hold the status of a
perpetrator under penal code. In our current legal framework, only natural and juridical
persons can be held criminally liable. As an AI system cannot act with intent
or negligence on its own, it cannot be subjected to criminal sanctions.
Therefore, AI is considered a tool—an "instrument-subject" that
facilitates the commission of a crime rather than being the one who commits it.
5.2. Responsibility
of the Programmer
Programmers who
design and code AI systems directly shape the algorithms and decision-making
capabilities of those systems. If the software contains code that enables or
encourages criminal behavior, the programmer may be held liable. Especially in
cases involving intentional programming errors, security vulnerabilities, or
insufficient oversight, criminal liability may arise either through intent or
negligence on the part of the developer.
5.3. Responsibility
of the User
An individual who
actively uses an AI system—whether as an employee or a private user—may bear
criminal liability if they act based on the system’s outputs or provide data
inputs that guide the system's behavior. If the user is in a position to
foresee the unlawful outcomes generated by the system and nevertheless fails to
intervene, they may be held liable for negligent conduct.
5.4. Responsibility
of the Owner or Producing Company
The liability of
companies that own or commercially distribute AI systems can be analyzed within
the framework of corporate fault and organizational negligence. In particular,
if due diligence is not exercised during development, or if the product is released
despite known risks, criminal liability may be attributed to the juridical
person or its executives.
5.5. Shared and
Joint Liability
In some instances,
criminal acts involving AI may give rise to collective responsibility rather
than individual liability. For example, when a harmful result is caused by the
combined effect of programmer error, user misuse, and a company’s failure to provide
adequate oversight, joint and several liability may apply. In such cases, not
only the identity of the perpetrator but also the degree of fault plays a
significant role in sentencing.
6. Penal Code
Implications of AI-Based Content Generation: A Case Study of Grok
The implications of
artificial intelligence in the field of penal code have become a pressing issue
across many legal systems worldwide. In particular, the emergence of large
language models like Grok has triggered intense debates surrounding legal
responsibility, freedom of expression, and hate speech in the context of AI-generated
content.
6.1. Recent
Developments Concerning Grok
6.1.1. First
Official Intervention in Türkiye: The Grok case dramatically exemplifies
the tension between technological advancement and legal accountability. The
exploitation of jailbreak vulnerabilities in Grok to produce content involving
hate speech, insults, and incitement to violence has exposed significant gaps
in traditional penal code frameworks. For the first time in Türkiye, an AI
chatbot faced access restrictions and a potential criminal investigation due to
content targeting the President, Atatürk, and religious values. This incident
serves as a socio-legal milestone, illustrating how broadly the definition of a
“perpetrator” may be stretched. The Ankara Chief Public Prosecutor’s Office
imposed an access ban on Grok, citing “insulting content” related to President
Erdoğan, Atatürk, and religious values. The Information and Communication
Technologies Authority (BTK) enforced the decision, restricting access with reference to around 50 flagged items. [1]
[2]
6.1.2. Poland’s
Complaint to the EU:
The Polish Minister
for Digital Affairs announced plans to refer the matter to the European
Commission, citing Grok’s generation of antisemitic and defamatory content
concerning political figures such as Prime Minister Donald Tusk. [3]
6.1.3. Content Removal Following Hate Speech
Allegations:
Grok also faced
backlash after antisemitic content praising Adolf Hitler was disseminated via
its X platform account. Following complaints by the Anti-Defamation League
(ADL), the developer company xAI announced the removal of such content. [4]
7. Evaluation in
Respect of Turkish Penal Code Application
The advancement and
increasing use of artificial intelligence technologies in Türkiye have
introduced new challenges in the realm of penal code. However, as of now, the
Turkish Penal Code (TPC) does not contain any specific provisions directly
addressing AI-related issues. This absence of regulation results in legal
uncertainties regarding the basis for determining criminal liability in
concrete cases involving AI.
7.1. Existing Legal
Framework and Its Limitations
The lack of
specific provisions on AI in the TPC implies that responsibility must be
assigned to human actors. Nevertheless, under the general provisions of the TPC
—particularly those concerning intentional and negligent offenses—individuals
such as developers, users, or system owners may be held criminally liable for
harmful or criminal acts committed by AI. For example, regulations concerning
cybercrimes (Articles 243–244 of the TPC) may be applied indirectly to
AI-related misconduct, although their scope remains narrow and limited.
7.2. Normative Gaps
and the Need for Regulation
Serious gaps remain
regarding who qualifies as the perpetrator in AI-related offenses, how
culpability is to be determined, the extent of AI's autonomous behavior, and
the scope of responsibility. These deficiencies conflict with the principles of
legal certainty and foreseeability and complicate the protection of victims'
rights.
In both academic
and practical circles in Türkiye, efforts are growing to develop penal code regulations
specific to AI. Priority areas should include:
- The introduction
of specific offense types related to AI-based crimes,
- Clarification of
the responsibilities of developers and users,
- Defining the
criminal liability of platform and system owners.
8. Conclusion
Acts committed
through artificial intelligence constitute a complex area that is not
adequately addressed under current penal code frameworks. Since AI cannot be
recognized as a perpetrator, liability is primarily concentrated on human
actors such as developers, users, and system owners. However, the absence of
specific provisions regarding this issue in the Turkish Penal Code leads to
legal gaps and uncertainty in practice.
Contemporary
examples—such as legal controversies surrounding AI-generated content by models
like Grok—demonstrate the need to reassess both the boundaries of freedom of
expression and the foundations of criminal liability. It is of great importance
for Türkiye to adopt new regulations specifically tailored to AI, in line with
international standards, that establish clear accountability and oversight
mechanisms.
In conclusion, the
innovations brought about by artificial intelligence necessitate normative
reforms in penal code, updates in judicial practices, and the adoption of
multidisciplinary approaches. Only through such developments can the benefits
of AI be harnessed safely and legal justice be effectively maintained.
Efe Öztürk, Legal Intern
References:
1. Turkish Penal
Code
2. Turkish
Civil Code
6. https://www.theguardian.com/technology/2025/jul/09/grok-ai-praised-hitler-antisemitism-x-ntwnfb
