Within the tapestry of human progress, a concept emerges that far exceeds the boundaries of mere technology—a notion that transcends the tangible and delves into the ethereal. At the nexus of innovation and imagination resides the enigmatic realm of Artificial Intelligence, a term too complex to be confined to a singular definition, encompassing an intricate fusion of technological prowess and the depths of human ingenuity. The European Union is attempting to keep control over the dreams and fears within the digital world’s realm. The aim of this article is to introduce the Digital Services Act and its provisions regarding freedom of expression and provide a future outlook on the content of the EU’s AI Act which is still under construction.

After the Internet became one of the most frequently used spaces for information gathering and sharing, the general international human rights treaties seemed to be too generic for protecting freedom of expression on-line[1]. In the European Union (EU), it was the Charter of Fundamental Rights (the Charter) that aimed to protect freedom of expression (besides the national constitutional protection provided by each Member State), especially in the media, and clarify the boundaries of this right amongst many in all Member States[2]. Even the consolidated version of the Charter functions as a dogmatic framework, that provides a bouquet of human rights without more specific determination of these terms. With the spread of the internet and the offering of more and more private to public services online, states and the EU have realized that the harmonization of existing domestic rules and EU-level regulation to protect fundamental human rights in the online realm seems necessary. The biggest concern was evolving around the freedom of expression on online platforms.[3] In 2022, the Digital Services (DSA)[4] and the Digital Markets Acts (DMA)[5] were proclaimed that is about to serve as regulatory frameworks for online platforms, in the case of DSA for their services and their online platforms, and in the case of DMA, the online competition. These two acts will enter into force in 2024. In the timeline of events, the General Data Protection Regulation (GDPR)[6] has appeared already in 2016, and it covers data protection of natural persons and online consumer protection for services and identifiers as well. Furthermore, there are some other legislations regulating the online space, but the most comprehensive ones are the DSA and the DMA, and the parts of the GDPR, too.

In this European patchwork of online regulatory schemes, continuous threats were revealed by the increased use of Artificial Intelligence (AI). For example, the creation of deep fake videos about the pope and politicians might seem like great jokes, however, they easily discredit public trust and question the relevance of public institutions. As a natural consequence of this situation, the EU showed interest in regulating AI separately and introducing a more detailed framework that deals with this very specific concept[7].

A dominant purpose of the DSA was to strengthen the protection of freedom of expression, and not solely to highlight the regulatory limitations but to try to emphasize this fundamental right[8]. However, it still states that it also aspires to determine the boundaries of this right to protect users from illegal content, hate speech, and so on[9]. Among the boundaries planned are for example the establishment of flagging systems, where content could be judged by subjective opinions and not necessarily on the grounds of objective illegality, which in fact disables predictability.

The importance of freedom of expression is revealed already in the Preamble of the document[10]. The DSA reflects on the Charter and treats its content as a dogmatic background specifically for big tech[11]. The requirement from large search engines and online platforms is to create procedures whose mechanisms are understandable for the platforms’ and engines’ users[12]. Apart from mentioning the term in the Preamble, four separate articles cover partially the freedom of expression from different angles.[13] For instance, Article 22 highlights if the platform intends to avoid responsibility and potential liability for the posts on its forum, it has to be cautious about removing illegal content[14]. However, the effectiveness of the articles in safeguarding this fundamental right is still questioned by many.

Article 14 in paragraph 4 directs intermediary service providers to act with proportionality and objectivity when it comes to restricting fundamental rights, especially freedom of expression and media pluralism[15]. While Article 22 forewarns the platforms and search engines that they need to heed the removal of illegal content if they aim to be exempt from liability for the content that raises the concern of abuse of power by these platforms. The Act describes the extent and way of the potential restrictions with more detail by steadily bearing the fundamental rights in mind and stating their importance explicitly in the text of the DSA. A new approach of the DSA was to concretize the penalty for unjust decisions, as previous regulations did not offer any frame for such fines or compensations. The potential fine can account for a maximum of 6% of the turnover from the preceding financial year[16]. This can be considered a significant amount; however, this rate ceiling does not provide a proper frame as a minimum fine is not determined by the DSA and does not ensure enforcement from the European Commission’s side.[17]

By regulating “very large platforms” (VLOPs) separately and with more precaution, the DSA ensures oversight and compliance with its regulations as they propose a greater systemic risk to fundamental rights than smaller platforms[18]. In the case of large platforms, a small mistake in the algorithm can create a snowball effect that could affect the users’ fundamental rights and as they have a significant number of profiles on their platforms, the violation hinders more people as well. Article 34 prescribes such risk assessment for these VLOPs to reveal systemic risks in their algorithmic systems where likelihood and severity need to be accounted for in their sector-specific services. Such risks include the probability of the appearance of illegal content, threat to fundamental rights of the Charter, and potential damage to the democratic electoral system[19]. The assessment has to aim towards minimizing the risk of over-blocking content as well as undermanaging the platform as the stake of damage in free speech on larger platforms is more crucial[20]. Article 86 confirms that systemic risk for VLOPs is much greater since even a small mistake in the algorithm can cause significant accumulated damages. Articles 63 and 81 intend to ensure avoidance of abuse of power and try to counterweigh the strength of these platforms[21]. As the article has shown, the DSA targets these large entities, in other words, VLOPs, as they pose a serious threat due to their market share and operational reach.

Platforms must establish a crisis protocol to attempt to safeguard the Charter’s fundamental rights, specifically freedom of expression[22]. Various control mechanisms, such as the internal complaint-handling system, are offered by the DSA to ensure the recovery of misguided decisions of content restrictions on major platforms. There has to be a “put-back” mechanism that secures room for appealing those harmful decisions[23]. The DSA incorporates three distinct control stages where procedures for the correction of enormous decisions are foreseen. Emphasis is laid on “internal complaint-handling systems”, which are established entities within or beside the company itself, whose purpose is to deal with users’ complaints about content moderating decisions made by the platform.[24] Most platforms already introduced opportunities for the users to appeal a decision and restoration rates are regarded as somewhat successful[25]. Apart from voluntary in-house solutions, establishing external bodies that are set to self-regulate is becoming more and more common. Such bodies, such as Facebook’s Oversight Board are not sufficient to ensure a user’s right to appeal as only the platform has the decisive power to refer problematic cases to the Oversight Board[26]. As a final channel to appeal faulty decisions, the Member State Courts provide the opportunity for users to sue the platforms for overblocking where courts tend to secure preliminary restoration, as financial risks tend to be more modest for a corporation[27]. However, such disputes are usually settled by the platforms, and only in a few cases reach a court verdict. One of the biggest lawsuits involving Facebook was the Cambridge Analytica scandal regarding the 2016 elections in the US. Facebook agreed to a settlement of $725 million in April 2023 and will provide compensation for the users with accounts between 2007 and December 2022[28]. However, this is rather the exception than the rule.

Apart from large online platforms’ content moderation practices, another aspect of digitalization threatens social media users. Artificial intelligence can influence fundamental rights such as the freedom of expression significantly by profiling content consumers with big data gathering and also by content removal and moderation[29]. Even before the DSA was proclaimed, the European Commission proposed a draft to regulate artificial intelligence on a larger scale and in a unified manner in 2021. The AI Act, when introduced, is expected to become the most comprehensive legal document regulating AI with an impact on third countries outside the Union as well since third parties and countries need to comply with the regulation if their content and system are planned to be made available within the EU[30]. However, still many people question the ability to regulate AI effectively in practice. As algorithms based on AI generate a filtering process according to their own standards and guidelines, artificial intelligence regulation could, regulated effectively, play a significant role in preventing misuse of power in content moderation to protect freedom of expression[31]. However, this could only be achieved by a strong regulation. The proposal for the AI Act incorporates the fundamental rights listed in the Charter and these rights are even mentioned in a separate section of the explanatory memorandum[32]. The Act prescribes required transparency for specific AI systems that interact with natural persons. In that manner, these natural persons had to be notified of the use of AI in their interactions, especially if they are using biometric face recognition or artificially generated content such as deepfakes (Article 52). This shall be exempt from application only where the law specifically authorizes the party to prevent criminal activities and with that, it contributes to freedom of expression[33]. The document would require systems to classify their activities according to risk, similarly to the DSA. To the original proposal, amendments were enclosed in 2023 June, and currently, the legislation is still under negotiation at the table of the European Commission and the member states. However, this is expected to be a rather lengthy procedure until the act is proclaimed as law[34].

The EU is clearly determined to regulate the online space and would have the intention to protect the fundamental rights of the Charter. However, the Union still aims to benefit from the use of AI and does not intend to overregulate the sector. With the risk tier systems, the threat of limiting the use of AI extensively could be resolved in theory, but both the DSA and the AI Act have to prove their practical effectiveness in the future. The DSA was accepted only last year and seemed to overregulate but in practice, it is expected to be inefficient to protect the implicated fundamental rights. However, the document is not in force yet, therefore, it is still uncertain what the practical effects will be. The AI Act is solely a proposal that is still under construction and up to negotiation, but the practical use of its content is yet questionable.

---

[1] Ling, Jiang and Yue, Zhang: Research on the Displacing Effect of the Internet on the Traditional Media. In. International Journal of Social Science and Humanity, Vol. 5, No. 7, July 2015, p.594.

[2] Equality and Human Rights Commission: What is the Charter of Fundamental Rights of the European Union? Updated on 2 August 2021. https://www.equalityhumanrights.com/en/what-are-human-rights/how-are-your-rights-protected/what-charter-fundamental-rights-european-union Accessed on 12 August 2023

[3] European Commission: The Digital Services Act: ensuring a safe and accountable online environment. 2022. In. European Commission. https://commission.europa.eu/strategy-and-policy/priorities-2019-2024/europe-fit-digital-age/digital-services-act-ensuring-safe-and-accountable-online-environment_en Accessed on 16 July 2023.

[4] European Union: Regulation (EU) 2022/2065 of the European Parliament and of the Council on a Single Market for Digital Services and amending Directive 2000/31/EC (Digital Services Act). 2022.

[5] European Union: Regulation (EU) 2022/1925 of the European Parliament and of the Council on contestable and fair markets in the digital sector and amending Directives (EU) 2019/1397 and (EU) 2020/1828 (Digital Markets Act). 2022.

[6] European Union: Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, and repealing Directive 95/46/EC (General Data Protection Regulation). 2016.

[7] Marchenko, Anna Y., Entin, Mark L.: Artificial Intelligence and Human Rights: What is the EU’s approach?. In. Digital Law Journal. Vol. 3. No. 3. Published on 30 September 2022. pp. 43-57, https://doi.org/10.38044/2686-9136-2022-3-3-43-57 p.43.

[8] European Publishers Council: The Digital Services Act must safeguard freedom of expression online. 2022. https://www.epceurope.eu/post/the-digital-services-act-must-safeguard-freedom-of-expression-online#:~:text=3%20min-,The%20Digital%20Services%20Act%20must%20safeguard%20freedom%20of%20expression%20online,between%20citizens%20online%20is%20curtailed. Accessed on 15 August 2023.

[9] European Union: Regulation (EU) 2022/2065 of the European Parliament and of the Council on a Single Market for Digital Services and amending Directive 2000/31/EC (Digital Services Act) – Preamble (9), Preamble (12). 2022.

[10] European Union: Regulation (EU) 2022/2065 of the European Parliament and of the Council on a Single Market for Digital Services and amending Directive 2000/31/EC (Digital Services Act) – Article 14, Article 34, Article 47, Article 91. 2022.

[11] European Union: Regulation (EU) 2022/2065 of the European Parliament and of the Council on a Single Market for Digital Services and amending Directive 2000/31/EC (Digital Services Act) – Preamble (3). 2022.

[12] European Union: Regulation (EU) 2022/2065 of the European Parliament and of the Council on a Single Market for Digital Services and amending Directive 2000/31/EC (Digital Services Act). 2022.

[13] European Union: Regulation (EU) 2022/2065 of the European Parliament and of the Council on a Single Market for Digital Services and amending Directive 2000/31/EC (Digital Services Act) – Article 14, Article 34, Article 47, Article 91. 2022.

[14] European Union: Regulation (EU) 2022/2065 of the European Parliament and of the Council on a Single Market for Digital Services and amending Directive 2000/31/EC (Digital Services Act) – Article 22. 2022.

[15] European Union: Regulation (EU) 2022/2065 of the European Parliament and of the Council on a Single Market for Digital Services and amending Directive 2000/31/EC (Digital Services Act) – Article 4. Paragraph 4. 2022.

[16] European Union: Regulation (EU) 2022/2065 of the European Parliament and of the Council on a Single Market for Digital Services and amending Directive 2000/31/EC (Digital Services Act) – Article 52. 2022.

[17] Wagner, Ben, Janssen Heleen: A First Impression of Regulatory Powers in the Digital Services Act. In. Verfassungsblog. January 2021. https://verfassungsblog.de/regulatory-powers-dsa/ Accessed on 15 August 2023

[18] Wagner, Ben, Janssen Heleen: A First Impression of Regulatory Powers in the Digital Services Act. In. Verfassungsblog. January 2021. https://verfassungsblog.de/regulatory-powers-dsa/ Accessed on 15 August 2023.

[19] European Union: Regulation (EU) 2022/2065 of the European Parliament and of the Council on a Single Market for Digital Services and amending Directive 2000/31/EC (Digital Services Act) – Article 34. 2022.

[20] Holznagel, Daniel: The Digital Services Act wants to “sue” Facebook over content decisions in private de facto courts. In Verfassungsblog. Juni 2021. https://verfassungsblog.de/dsa-art-21/ Accessed on 14 August 2023.

[21] European Union: Regulation (EU) 2022/2065 of the European Parliament and of the Council on a Single Market for Digital Services and amending Directive 2000/31/EC (Digital Services Act) – Article 63. Article 81. 2022.

[22] European Union: Regulation (EU) 2022/2065 of the European Parliament and of the Council on a Single Market for Digital Services and amending Directive 2000/31/EC (Digital Services Act) – Article 47. 2022.

[23] Holznagel, Daniel: The Digital Services Act wants to “sue” Facebook over content decisions in private de facto courts. In Verfassungsblog. Juni 2021. https://verfassungsblog.de/dsa-art-21/ Accessed on 14 August 2023.

[24] Wagner, Ben, Janssen Heleen: A First Impression of Regulatory Powers in the Digital Services Act. In. Verfassungsblog. January 2021. https://verfassungsblog.de/regulatory-powers-dsa/ Accessed on 15 August 2023

[25] Holznagel, Daniel: The Digital Services Act wants to “sue” Facebook over content decisions in private de facto courts. In Verfassungsblog. Juni 2021. https://verfassungsblog.de/dsa-art-21/ Accessed on 14 August 2023.

[26]Holznagel, Daniel: The Digital Services Act wants to “sue” Facebook over content decisions in private de facto courts. In Verfassungsblog. Juni 2021. https://verfassungsblog.de/dsa-art-21/ Accessed on 14 August 2023.

[27] Holznagel, Daniel: The Digital Services Act wants to “sue” Facebook over content decisions in private de facto courts. In Verfassungsblog. Juni 2021. https://verfassungsblog.de/dsa-art-21/ Accessed on 14 August 2023.

[28] Campisi, Natalie: Facebook Privacy Settlement Payouts: Here’s how much money you’ll get. In. Forbes. Updated on 13 September 2023. https://www.forbes.com/advisor/legal/facebook-class-action-lawsuit-settlement/#:~:text=About%2017.7%20million%20Facebook%20users,lawyers%20involved%20in%20the%20suit. Accessed on 20 September 2023.

[29] Kolarević, Emina: The Influence of Artificial Intelligence on the Right of Freedom of Expression. In. Pravo. 2022. Vol 39. No. 1. p.111. https://doi.org/10.5937/ptp2201111K Accessed on 13 August 2023.

[30] Sathe, Madan, Ruloff, Karl: The EU regulation for artificial intelligence is coming. What does it mean for you and your business in Switzerland?. In. EY. Published on 28 July 2023. https://www.ey.com/en_ch/forensic-integrity-services/the-eu-ai-act-what-it-means-for-your-business Accessed on 20 September 2023.

[31] Kolarević, Emina: The Influence of Artificial Intelligence on the Right of Freedom of Expression. In. Pravo. 2022. Vol 39. No. 1. p.119. https://doi.org/10.5937/ptp2201111K Accessed on 13 August 2023.

[32] European Union: Proposal for a Regulation of the European Parliament and of the Council Laying Down Harmonised Rules on Artificial Intelligence and Amending Certain Union Legislative Acts. – Artificial Intelligence Act. – 3.5. Fundamental Rights. https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:52021PC0206

[33] European Union: Proposal for a Regulation of the European Parliament and of the Council Laying Down Harmonised Rules on Artificial Intelligence and Amending Certain Union Legislative Acts. – Artificial Intelligence Act. – Article 52. https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:52021PC0206

[34] Feingold, Spencer: The European Union’s Artificial Intelligence Act – explained. In. World Economic Forum. 2023. https://www.weforum.org/agenda/2023/06/european-union-ai-act-explained/ Accessed on 16 August 2023.

---

Dorina BOSITS is a law student at the Széchenyi István University of Győr, Hungary, and an international finance and accounting graduate of the University of Applied Sciences of Wiener Neustadt, Austria. The main area of her research includes freedom of speech, digitalization, data protection, and financial law. She is a student at the Law School of MCC and a member of ELSA Győr.