by

Yasmine Ouali

CyJurII Scholar

on 9 July 2026

 PDF Available

Introduction

Over the past decade, the world has witnessed a radical transformation in the production of knowledge and the dissemination of information, driven by the rapid development of artificial intelligence (AI), particularly generative AI systems. These systems are capable of producing text, images, audio recordings, and videos with a high degree of realism, making the distinction between human-authored and machine-generated content increasingly difficult to identify. Their widespread use has fundamentally reshaped the digital sphere and the contemporary public square. Algorithms are no longer confined to the role of technical intermediaries in transmitting information; rather, they have become influential actors in producing, structuring, directing, and shaping public opinion, as well as influencing social and political decision-making. Consequently, AI may be regarded as a new stage in the evolution of mass communication, extending beyond the traditional digitisation of content. ¹

At the same time, this technological development has raised significant legal and human rights questions concerning the future of freedom of expression in the digital environment. While AI expands access to information and creates new opportunities for content creation and communication, it also gives rise to serious concerns regarding disinformation, deepfakes,

AI-generated hate speech, digital defamation, and algorithmic manipulation of public opinion. These risks directly affect the rights protected under article 19 of the Universal Declaration of Human Rights 1948 and article 19 of the International Covenant on Civil and Political Rights 1966, both of which guarantee the right to freedom of opinion and expression, including the right to seek, receive, and impart information and ideas without unlawful interference.²

The gravity of these concerns increases where intelligent systems generate unlawful or harmful content without direct human intervention. This development poses a significant challenge to traditional rules of criminal liability. Classical criminal law generally assumes the existence of a natural or legal person to whom criminal conduct may be attributed, together with the concurrent fulfilment of the actus reus and mens rea of the offence. AI systems, however, possess neither legal personality nor independent will or cognition in the traditional legal sense. This raises complex questions concerning the attribution of liability when criminal acts are committed through AI systems. Should liability rest with the programmer, developer, corporate owner, deployer, platform, or end-user? Or does the emergence of generative AI require a rethinking of the traditional concepts of criminal accountability?

This debate is no longer merely theoretical. It has become the subject of growing legislative and judicial attention at the international and comparative levels. The European Union adopted the first comprehensive and binding legal framework on artificial intelligence through Regulation (EU) 2024/1689, known as the Artificial Intelligence Act (‘EU AI Act’).³ The Act adopts a

risk-based model, classifying AI systems according to the degree of risk they pose to fundamental rights, safety, and public health, and imposing graduated obligations as the level of risk increases.

The EU AI Act gives particular importance to the protection of fundamental rights, including freedom of expression, privacy, and non-discrimination. It prohibits certain AI practices that create unacceptable risks, such as social scoring and certain forms of biometric surveillance, while also imposing transparency and disclosure obligations for AI-generated or AI-manipulated content, including deepfakes and public-facing synthetic content.⁴

In parallel, the Council of Europe adopted the Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law in 2024.⁵ This Convention is the first legally binding international treaty addressing AI, human rights, democracy, and the rule of law. It seeks to ensure that the design, development, and deployment of AI systems remain consistent with international human rights standards, democratic safeguards, and the principles of legality and accountability.

Prior to these developments, UNESCO adopted its Recommendation on the Ethics of Artificial Intelligence in 2021.⁶ The Recommendation emphasises transparency, accountability, human oversight, fairness, non-discrimination, and the preservation of human dignity as foundational principles for the international governance of AI.

At the Arab regional level, specialised legislative frameworks on AI remain relatively limited when compared with international developments. However, several national initiatives have emerged to establish ethical and governance principles for this technology. Jordan was among the first Arab states to adopt a structured national framework through the National Charter for AI Ethics, issued by the Ministry of Digital Economy and Entrepreneurship in 2020.⁷ The Charter draws upon Jordan’s National Artificial Intelligence Policy and international ethical principles, emphasising justice, non-discrimination, transparency, data protection, and algorithmic accountability. Similarly, Saudi Arabia and the United Arab Emirates have adopted national AI strategies and ethical frameworks designed to promote trustworthy AI and create an enabling regulatory environment.⁸

In light of these developments, it is necessary to examine the relationship between AI and freedom of expression on the one hand, and criminal liability for AI-generated acts and content on the other. These are among the most pressing legal issues of the contemporary digital age, particularly in the absence of a unified Arab legal framework capable of balancing the protection of fundamental rights with effective accountability for harms arising from the use of AI systems.

Main Research Question

In view of the expanding role of AI in producing and directing digital content, and the increasing risks associated with disinformation, deepfakes, hate speech, and cybercrime, the central research question is:

To what extent can freedom of expression in the digital environment be reconciled with criminal accountability for harms and offences arising from the use of AI systems, without undermining fundamental rights or stifling technological innovation?

This central question gives rise to several subsidiary questions:

1. What is the impact of AI systems on the exercise of freedom of expression in the digital sphere?

2. Are traditional rules of criminal liability sufficient to address crimes committed through AI?

3. What are the boundaries of liability for developers, users, and digital service providers in relation to AI outputs?

4. To what extent have comparative frameworks, particularly the European model, succeeded in balancing fundamental rights protection with legal accountability?

5. What avenues exist for developing Arab and Tunisian legal frameworks capable of addressing the challenges created by AI?

Artificial Intelligence and Freedom of Expression

1. Artificial Intelligence as a New Tool for Exercising Freedom of Expression

Freedom of expression is a central pillar of the contemporary human rights system. Article 19 of the ICCPR protects the right to hold opinions without interference, as well as the freedom to seek, receive, and impart information and ideas of all kinds, regardless of frontiers, and through any medium of one’s choice.⁹ This provision demonstrates that freedom of expression is not merely a negative right against state interference, but also a positive entitlement to use available means of communication to generate and disseminate ideas.

The Tunisian Constitution adopts a similar approach. Article 37 of the Constitution of the Republic of Tunisia 2022 provides that ‘the freedoms of opinion, thought, expression, information and publication shall be guaranteed’ and that ‘no prior censorship shall be exercised over these freedoms’.¹⁰ This places freedom of expression within the constitutional framework of rights and freedoms and confirms that any restriction must comply with strict requirements of legality, necessity, proportionality, and protection of public order.

However, the emergence of generative AI has reframed freedom of expression in a new and unconventional context. Expression is no longer exclusively tied to direct human action. It is increasingly mediated through algorithmic systems capable of producing, modifying, and reformulating content with varying degrees of autonomy. Large language models (‘LLMs’), such as ChatGPT, Gemini, and Claude, illustrate this structural transformation in the production of discourse. Users can now generate large volumes of text, images, and audiovisual material within seconds, often with a quality that simulates or exceeds human production in certain technical respects.

This development raises an important legal question concerning the nature of AI-generated outputs. Should such outputs be regarded as indirect human expression, or do they constitute autonomous machine-generated content falling outside the constitutional protection of freedom of expression?

One view treats AI as an extension of the expressive tools available to human beings. On this approach, AI-generated content may be attributed to the user who directed the system, formulated the prompt, and determined the communicative purpose of the output. Where the content reflects a conscious human intention, it may fall within the protection of freedom of expression.

This approach is broadly consistent with expansive interpretations of freedom of expression in American constitutional scholarship, particularly in relation to digital expression and the use of technological media for content creation. Under this view, constitutional protection does not depend upon the medium used, but upon the communicative substance of the expression and its connection to an identifiable human actor. AI therefore does not alter the legal nature of expression; it alters the means through which expression is produced.

A different approach, more common in recent European scholarship, is more cautious. It argues that AI outputs should not automatically be treated as expression in the traditional legal sense, particularly where the element of direct human will or intent is absent or weakened. On this view, constitutional protection remains attached to natural and legal persons, as they are the recognised subjects of rights and obligations. Algorithms remain technical instruments without legal personality, consciousness, or legally recognised intention.

The challenge therefore lies not only in determining whether AI is a new vehicle of expression, but also in rethinking the concept of expression within the digital ecosystem. The central question becomes whether an act of expression can be attributed to a system that has no legal consciousness, particularly where the final wording, image, or audiovisual output is generated without direct human intervention.

This issue is particularly significant in journalism, political communication, and electoral discourse. AI-generated content can influence public opinion while obscuring its source, authorship, or authenticity. This creates a paradox for freedom of expression: AI expands expressive capacity while simultaneously increasing the risk of manipulation, distortion, and non-transparent influence.

Accordingly, AI is no longer merely an ancillary tool for exercising freedom of expression. It is a structural component that reshapes the scope, limits, and mechanisms of that freedom. This requires a renewed reading of constitutional and human rights principles in light of contemporary digital transformations.

2. Algorithmic Censorship and the Boundaries of Legality

In recent years, the digital environment has undergone a profound transformation in the regulation and dissemination of content. The moderation of public discourse is no longer carried out solely through direct human intervention. It increasingly relies on AI systems and algorithmic tools to scan, classify, restrict, remove, or downrank digital content in an automated or semi-automated manner. This process is commonly referred to as algorithmic censorship or automated content moderation.

These systems typically rely on machine learning and natural language processing techniques to detect content classified as harmful or unlawful, including hate speech, terrorist content, incitement to violence, disinformation, misinformation, and sexually explicit or violent material. However, the growing reliance on automated systems to make decisions affecting expression raises serious concerns regarding legality, transparency, due process, and access to remedies.

From a human rights perspective, freedom of expression may be restricted only where the restriction satisfies the requirements of legality, legitimate aim, necessity, and proportionality.¹¹ Algorithmic censorship risks undermining these requirements because users are often not informed of the precise criteria used to remove, restrict, or downrank their content. This may result in an ‘invisible restriction’ of freedom of expression, where content is suppressed without a reasoned decision, adequate explanation, or effective appeal mechanism.

UNESCO’s Recommendation on the Ethics of Artificial Intelligence warns against the unregulated use of AI in contexts affecting human rights and public discourse. It emphasises that AI systems must be subject to transparency, human oversight, accountability, and mechanisms for redress.¹² These principles are particularly important where automated moderation systems affect the exercise of freedom of expression.

The EU AI Act also reflects this concern. Article 50 imposes transparency obligations on providers and deployers of certain AI systems, particularly where content is generated, manipulated, or presented in a manner that may mislead users.¹³ This represents a legislative attempt to regulate the interface between AI technologies and freedom of expression within a broader risk-management framework.

The problem, however, extends beyond transparency. Automated moderation decisions are often produced by complex probabilistic systems. They are not necessarily based on a direct legal assessment of the content, but on statistical estimations of whether the content is likely to violate platform policies or legal rules. This raises a fundamental question about compatibility with the principle of legality, which requires restrictions on rights and freedoms to be clear, foreseeable, and established by law.

Algorithmic censorship therefore reshapes the concept of legality in the digital domain. The power to assess the lawfulness of content may shift from judges or competent administrative authorities to private technical systems that rely on opaque models. This produces what may be described as the privatisation of freedom of expression decisions: the transfer of regulatory authority over fundamental rights to private actors who are not subject to the same degree of public legal oversight.

The principal challenge posed by algorithmic censorship is not merely technical effectiveness, but compatibility with constitutional and human rights principles. These include legality, proportionality, transparency, due process, and the right to an effective remedy.

3. Deepfakes and the Threat to Freedom of Expression

Deepfakes are among the most significant and dangerous applications of generative AI. They can produce highly realistic audio, visual, and audiovisual content that is difficult to distinguish from authentic material. Under the EU AI Act, deepfakes are understood as AI-generated or AI-manipulated content resembling existing persons, objects, places, entities, or events and falsely appearing to be authentic or truthful.¹⁴

Deepfakes create a dual legal dilemma. On one hand, they may be used as a form of digital expression, including for artistic, satirical, educational, journalistic, or research purposes. On the other hand, they may seriously harm the rights of individuals and groups, particularly where they are used without disclosure or outside a legitimate context.

The risks associated with deepfakes include:

1. political or digital impersonation through fabricated audio or video recordings attributed to public figures;

2. falsification of journalistic discourse and erosion of public trust in the media;

3. interference with electoral and democratic processes through targeted disinformation; 4. damage to reputation through defamatory synthetic content; and

5. digital blackmail or coercion using fabricated sensitive material.

From a legal standpoint, the problem is not limited to the misleading nature of deepfake content. It also concerns the difficulty of applying traditional rules of criminal liability where multiple actors are involved, including the user, developer, platform, and provider of the AI system.

The EU AI Act addresses this risk by imposing disclosure obligations in relation to AI-generated or AI-manipulated content. Article 50 requires certain AI-generated outputs, including deepfakes, to be disclosed in a clear and distinguishable manner.¹⁵ This obligation reflects two important principles: transparency regarding the provenance of digital content, and protection of users against deception.

The regulation of deepfakes is therefore closely connected to freedom of expression. The law does not seek to impose a categorical prohibition on synthetic content. Rather, it seeks to regulate the conditions of production and dissemination in order to prevent malicious use while preserving legitimate forms of artistic, journalistic, research, and political expression.

Deepfakes thus present more than a technical challenge. They raise fundamental legal questions concerning truth, authenticity, reputation, democratic integrity, and the adequacy of traditional criminal and media law rules in the age of generative AI.

Criminal Liability for Artificial Intelligence

1. The Crisis of Criminal Attribution

Traditional criminal liability rests upon two essential elements: the actus reus, consisting of the prohibited conduct, consequence, and causal link; and the mens rea, consisting of criminal intent or negligence. Without these elements, criminal liability cannot ordinarily be established.

The insertion of AI systems into the chain of criminal conduct disrupts this traditional structure. AI systems do not possess independent legal will, consciousness, or criminal intent, even where they operate semi-autonomously through machine learning or generative models. This creates a gap in the traditional chain of attribution between the criminal act and the legally responsible actor.

The central question is therefore which person or entity may be held criminally liable where an offence is committed through or with the assistance of an AI system. Potentially responsible actors may include:

1. the programmer who designed the algorithm;

2. the developer or corporate entity responsible for the system;

3. the deployer or operator of the system;

4. the end-user who issued the prompt or instruction;

5. the data provider whose material contributed to the training of the model; and

6. the digital platform that enabled the publication or dissemination of the content.

This multiplicity of actors gives rise to what may be described as the diffusion of criminal attribution. Traditional rules of liability may struggle to identify the actor with sufficient control, knowledge, intention, or negligence to justify criminal responsibility.

2. Contemporary Jurisprudential Trends

A. The Human Liability Approach

The first and dominant approach maintains that humans remain responsible for AI-related offences. This view is deeply rooted in classical criminal law. It proceeds from the principle that criminal liability can attach only to a natural or legal person capable of legal responsibility, and that AI is a sophisticated technical tool rather than an autonomous legal subject.

Under this approach, liability may be distributed according to fault, foreseeability, and effective control:

1. Programmer liability may arise where a design defect, foreseeable vulnerability, or negligent coding practice contributes to criminal harm.

2. Developer or corporate liability may arise where the entity responsible for the system fails to manage risks, comply with safety standards, or implement appropriate safeguards.

3. User liability may arise where the user intentionally or negligently directs the system to produce unlawful or harmful content.

4. Platform liability may arise where a digital service provider fails to act after receiving notice of unlawful content or fails to comply with applicable regulatory obligations.

This approach reflects the principle of personal criminal responsibility. There can be no crime and no punishment without personal liability, fault, and legally attributable conduct. This principle remains a safeguard against strict or objective liability.

3. The Dilemma of Legal Personality for Artificial Intelligence

A second, more controversial approach proposes granting AI systems a distinct legal status, sometimes referred to as ‘electronic personhood’. This idea appeared in European Parliament debates on robotics and intelligent systems in 2017.¹⁶ However, it has not developed into binding law and has been widely criticised.

The principal objections are threefold. First, electronic personhood is difficult to reconcile with the requirement of mens rea. AI systems cannot form criminal intent, moral blameworthiness, or consciousness in the legal sense. Secondly, attributing liability to AI may shield the actual human actors who designed, deployed, controlled, or profited from the system. Thirdly, it conflicts with the philosophy of criminal law, which is grounded in deterrence, culpability, and the personal nature of punishment.

For these reasons, the prevailing position in comparative legislation rejects the attribution of criminal legal personality to AI. Instead, it favours adapting existing liability rules to capture the conduct of human and corporate actors involved in the design, deployment, operation, and use of AI systems.

4. Manifestations of Criminal Liability for AI-Generated Content

A. Hate Speech

Where an AI system is used to generate content involving incitement to hatred, discrimination, or violence, criminal liability should generally attach to the user who directed the system toward producing or disseminating the unlawful content. The user may be regarded as the intellectual author of the content, particularly where the prompt, intention, and publication are attributable to that person.

Liability may also extend to developers or platforms where they knowingly facilitate unlawful use, fail to implement reasonable safeguards, or ignore foreseeable risks. However, such liability must be carefully limited to avoid imposing excessive burdens on innovation or lawful expression.

B. Defamation and Libel

Where AI generates false information that harms the honour or reputation of an identifiable person, several actors may potentially bear responsibility. The user may be liable as the person who prompted, selected, edited, or published the defamatory content. A digital platform may be liable where it fails to remove unlawful content after receiving notice, depending on the applicable legal framework. A developer may be liable where gross negligence in system design or risk mitigation can be established.

This multi-actor model reflects the layered structure of digital harm. Liability should be assessed according to knowledge, control, foreseeability, and the causal connection between the actor’s conduct and the resulting harm.

C. Deepfakes

Deepfakes represent one of the most complex forms of AI-related criminal conduct. They may cause serious harm to reputation, privacy, security, and democratic trust. Criminal liability may arise where deepfakes are used for defamation, fraud, extortion, electoral manipulation, harassment, or threats to public safety.

The EU AI Act’s disclosure obligations indicate a broader shift in criminal policy from purely reactive punishment to prevention, transparency, and risk management. This preventive model may provide useful guidance for Arab and Tunisian legal systems seeking to regulate AI-generated harms while preserving legitimate expression.

Comparative Legislative Experiences

1. The European Union

The EU AI Act is the first comprehensive legislative framework regulating AI through a risk-based approach. It establishes several core principles, including transparency, accountability, human oversight, risk management, and the prohibition of certain unacceptable-risk AI practices.¹⁷

The Act imposes specific obligations on several categories of AI systems, including generative AI systems, automated conversational systems, deepfake technologies, and high-risk systems deployed in sensitive areas. Its significance lies not only in regulating AI as a technology, but also in connecting AI governance to the protection of fundamental rights.

2. The Council of Europe

The Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law links AI governance directly to states’ obligations under human rights law.¹⁸ It focuses on freedom of expression, privacy, equality, fair trial rights, democratic processes, and the rule of law. It represents the first binding international attempt to situate AI regulation within the human rights paradigm.

3. UNESCO

UNESCO’s Recommendation on the Ethics of Artificial Intelligence provides a non-binding but highly influential normative framework.¹⁹ It affirms that AI must be developed and used in accordance with human dignity, human rights, transparency, accountability, fairness, sustainability, and human oversight. Although not legally binding, it has become an important reference point for national and regional AI governance.

Arab Experiences in Regulating Artificial Intelligence and Protecting Fundamental Rights

Despite the absence of comprehensive Arab regional legislation comparable to the EU AI Act, several Arab states have made progress in AI governance through national strategies, ethical charters, and regulatory frameworks. These initiatives seek to balance technological innovation with the protection of fundamental rights.

1. Jordan: The Ethical Approach to Artificial Intelligence

Jordan was among the first Arab states to establish a structured national framework for AI governance. In 2020, the Ministry of Digital Economy and Entrepreneurship adopted the National Charter for AI Ethics.²⁰ The Charter affirms the principles of transparency, fairness, non-discrimination, privacy, personal data protection, and accountability for algorithmic decisions. It reflects a policy orientation that links AI development to respect for fundamental rights.

2. Saudi Arabia: Principles of AI Ethics

The Saudi Data and Artificial Intelligence Authority adopted the Principles and Controls of AI Ethics, updated in 2023.²¹ This framework is based on core principles including fairness, non-discrimination, privacy, human dignity, transparency, explainability, human oversight, accountability, and risk management. It also adopts a risk-based classification of AI systems, aligning in part with the European regulatory approach.

These principles are particularly relevant to freedom of expression because they emphasise the prevention of algorithmic bias, the reviewability of decisions affecting individuals, and the protection of fundamental rights against automated interference.

3. United Arab Emirates: Responsible AI Governance

In 2024, the United Arab Emirates adopted the Charter for the Development and Use of Artificial Intelligence.²² The Charter establishes principles for the responsible deployment of AI, including human oversight, privacy protection, mitigation of algorithmic bias, safety, legal compliance, and trust in smart technologies. The UAE has also adopted policy positions supporting international cooperation on AI governance.²³

Although these initiatives do not yet amount to detailed statutory regulation of freedom of expression comparable to the European model, they establish important principles of transparency, accountability, and human oversight. These principles are directly relevant to the protection of digital rights and the prevention of AI-enabled disinformation and content manipulation.

Overall, Arab experiences remain largely grounded in soft law, strategic frameworks, and ethical guidance rather than detailed binding legislation. Nevertheless, the Jordanian, Saudi, and Emirati models represent important steps toward developing legal and regulatory ecosystems capable of addressing freedom of expression and liability challenges arising from AI.

Conclusion

The rapid development of artificial intelligence has produced profound changes in the digital sphere and in the production and circulation of information. AI has created new opportunities for the exercise of freedom of expression, expanded access to knowledge, and broadened public participation. These developments are consistent with the guarantees contained in article 19 of the ICCPR and other international and domestic instruments.

However, these opportunities are accompanied by serious legal risks, including disinformation, deepfakes, hate speech, digital defamation, manipulation of public opinion, and offences that threaten legal certainty and fundamental rights. The central issue is not the existence of AI itself, but the manner in which it is designed, deployed, regulated, and used.

AI systems, despite their ability to generate seemingly autonomous outputs, remain incapable of possessing legal will or criminal intent under traditional criminal law. Directly attributing criminal liability to AI systems would conflict with foundational principles of criminal law, including legality, culpability, and the personal nature of punishment.

Comparative legal developments, particularly the EU AI Act, the Council of Europe Framework Convention, and UNESCO’s Recommendation, indicate that the prevailing international trend does not favour recognising independent legal personality for AI. Rather, it favours transparency, accountability, human oversight, risk management, and the distribution of responsibility among the human and corporate actors involved in designing, developing, deploying, and using AI systems.

The true challenge for Tunisian and Arab legal systems is therefore not merely to keep pace with technological change, but to develop a coherent legal framework that protects freedom of expression while ensuring accountability for AI-related harms. Such a framework must be grounded in international human rights standards, legal certainty, proportionality, and effective remedies.

Recommendations

In light of the foregoing analysis, the following recommendations may be proposed:

1. Adopt a specialized national legislative framework on AI. Tunisian and Arab legislators should establish general principles governing the development, deployment, and use of AI systems, drawing upon comparative best practices, particularly the EU AI Act.

2. Integrate AI-related criminal liability rules into penal legislation. Legislation should clarify the criteria for attributing liability to developers, deployers, users, operators, and digital platforms according to effective control, foreseeability of harm, knowledge, and fault.

3. Impose statutory obligations of algorithmic transparency. Digital service providers and AI system operators should be required to disclose AI-generated or AI-manipulated content, especially in relation to synthetic media and deepfakes.

4. Adopt a specific legal framework for deepfakes. The malicious use of deepfakes for defamation, fraud, electoral interference, blackmail, harassment, or threats to public safety should be criminalized, while preserving legitimate artistic, journalistic, academic, and satirical uses.

5. Protect freedom of expression against algorithmic censorship. Individuals should have the right to know why their content has been removed, restricted, or downranked, and should have access to human review and effective appeal mechanisms.

6. Establish independent oversight mechanisms for AI governance. Independent authorities should monitor compliance with statutory and ethical standards, investigate violations, and ensure that AI systems respect fundamental rights.

7. Support specialized legal and judicial training. Judges, prosecutors, lawyers, law enforcement officials, and forensic experts should receive training in AI, digital evidence, cybercrime, and algorithmic accountability.

8. Strengthen international and regional cooperation. AI-related crimes are often transboundary. States should cooperate in evidence gathering, technical expertise, legislative harmonization, and cybercrime enforcement.

9. Integrate AI ethics into public policy. Public policy should reflect the principles endorsed by UNESCO, including human dignity, non-discrimination, justice, transparency, accountability, and human oversight.

10.  Encourage specialized Arab legal research on AI. Arab legal scholarship should be strengthened to develop jurisprudential and legislative solutions that reflect regional legal systems, linguistic contexts, and social realities.

An effective legal architecture for AI should not be based on fear of technology or excessive restriction. Rather, it should be grounded in responsible governance, protection of rights and freedoms, legal accountability, and trust in the digital ecosystem.

References

International Instruments and Legislation

1. Universal Declaration of Human Rights 1948.

2. International Covenant on Civil and Political Rights 1966.

3. Constitution of the Republic of Tunisia 2022.

4. Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonized rules on artificial intelligence [2024] OJ L.

5. Council of Europe, Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law CETS No 225, 17 May 2024.

6. UNESCO, Recommendation on the Ethics of Artificial Intelligence (2021).

7. European Parliament, Civil Law Rules on Robotics: European Parliament Resolution of 16 February 2017 with Recommendations to the Commission on Civil Law Rules on Robotics (2015/2103(INL)) [2018] OJ C252/25.

Secondary Sources

1. Bayah Abd El Kader, ‘A Legal Approach to the Provisions of the Council of Europe Framework

2. Convention on Artificial Intelligence, Human Rights, Democracy and the Rule of Law’ (2026) 11(1) Algerian Journal of Legal and Political Sciences 311.

3. Badi Ben Abbas and Hatem Ben Jemaâ, The Principle of Proportionality in the Jurisprudence of the Tunisian Court of Cassation (International IDEA 2022).

4. Gyeonggi Lim, Generative Artificial Intelligence: What It Is, What It Is Not, and What It Could Mean for the United Nations (United Nations, 18 July 2023).

5. Jordan Ministry of Digital Economy and Entrepreneurship, National Charter for Artificial Intelligence Ethics (2020).

6. Majed Nasser Said Qadah, Criminal Legal Personality: A Comparative Jurisprudential Study with Law (Master’s thesis, An-Najah National University 2019).

7. Mohamed Ali Bouattia, ‘Artificial Intelligence Crimes’ (2025) 9(2) Qays Journal of Human and Social Studies 533.

8. Mohammed Ibrahim Mohammed Al-Amiri and Abdelilah Mohammed Salem Al-Nuwaiseh, ‘The Legal Foundations of Criminal Liability for Artificial Intelligence’ (2025) 4 Journal of Law, Kuwait University.

9. Mona Mohamed Al-Atrees Al-Desouky, ‘Crimes of Artificial Intelligence Technologies and Independent Electronic Legal Personality: A Comparative Study’ (2022) 12(81) Journal of Legal and Economic Research 1132.

10. Paweł Nowik, ‘Electronic Personhood for Artificial Intelligence in the Workplace’ (2021) 42 Computer Law & Security Review 105584.

11. Saudi Data and Artificial Intelligence Authority, Principles and Controls of AI Ethics RAI-SA-NA-PCAESXX-2023 (1 September 2023).

12. Selina Saâdoun, La personnalité juridique de l’intelligence artificielle (Master’s thesis, University Mouloud Mammeri of Tizi Ouzou 2021–2022).

13. United Arab Emirates, Charter for the Development and Use of Artificial Intelligence (10 June 2024).

14. Yasser Abdel Moneim Mohammed, ‘Artificial Intelligence and Its Role in Criminal Law’ (2025) 73(5) Journal of the Iraqia University.

15. Zahia Rabti, ‘Freedom of Expression in the Digital Age from the Perspective of International Law between Enshrinement and Restriction’ (2023) 8(2) Afak Journal of Sciences 492.

Footnotes

¹ Gyeonggi Lim, Generative Artificial Intelligence: What It Is, What It Is Not, and What It Could Mean for the United Nations (United Nations, 18 July 2023).

² Universal Declaration of Human Rights 1948, art 19; International Covenant on Civil and Political Rights 1966, art 19.

³ Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonized rules on artificial intelligence [2024] OJ L.

⁴ ibid art 50.

⁵ Council of Europe, Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law CETS No 225, 17 May 2024.

⁶ UNESCO, Recommendation on the Ethics of Artificial Intelligence (2021).

⁷ Jordan Ministry of Digital Economy and Entrepreneurship, National Charter for Artificial Intelligence Ethics (2020).