12.2.1 Copyright Law

The enforcement of copyright law online is likely one of the most characteristic areas where a united position gradually diverged into substantial differences between US and EU perspectives. It cannot be denied that the EU’s original stance on this issue, as expressed through the E-Commerce Directive (Directive 2000/31), which implemented a safe harbour protecting online intermediaries from liability, was heavily influenced by the American Digital Millennium Copyright Act, signed in 1998. It is worth noting that the regulation establishes the principle of prohibiting general internet monitoring, which remains in force today. However, as discussed later, efforts to create mechanisms of a priori control of the content have multiplied.

However, Europe’s position shifted, primarily through judicial activism, particularly in its strict interpretation of the safe harbour under ‘passive role’ theory.Footnote ⁵ The focus on a strict notice and takedown procedure inherently carries the danger of a ‘broad censorial attitude’,Footnote ⁶ with adverse effects on freedom of expression.Footnote ⁷

Furthermore, Telekabel jurisprudence paved the way for a new mode of the online enforcement of copyright law at the expense of freedom of expression.Footnote ⁸ In this case, the European Court of Justice (ECJ) admitted that an injunction against an internet service provider requesting the intermediary to implement filtering measures to block access to a website infringing copyright does not violate freedom of expression. This stance can be seen as a form of censorship, as it blocks user access to online information. However, the Court takes care to add two fundamental safeguards: the injunction should be effective, and it should never have the effect of blocking access to legal content.

One of the most significant and widely discussed interventions of the EU legislator regarding freedom of expression for copyright protection is Article 17 of Directive 2019/790. The purpose of Article 17 is to compel major internet service providers (modelled after Alphabet’s YouTube platform) to install filters that automatically block the uploading of copyright-infringing content. However, after extensive discussions and political negotiations, the term ‘filter’ itself does not appear in the final text of the Article. Article 17(4) only states that online service providers must make their best efforts to block this content; otherwise, they cannot rely on the old safe harbour mechanism.

Similar to the Telekabel case, it is evident how this regulation directly impacts freedom of expression. As a form of censorship, providers are required to block any attempts by users to communicate. This issue was immediately brought to the attention of the ECJ, which issued its ruling in the EU Commission vs. Poland case. The Court considered that Article 17(4) should indeed be interpreted as a restriction on freedom of expression. Still, it justified this restriction by citing the legitimate interest of protecting the rights holders. The Court outlined multiple safeguards: a system of internal and swift complaint mechanisms against abuse, limiting the filter to identified and protected works of intellectual property, restricting the filter’s algorithm to obvious situations of plagiarism, and protecting fundamental rights through exceptions such as short citations and parody.

12.2.2 Protection of Reputation

The ECJ’s landmark case, Eva Glawischnig-Piesczek v. Facebook,Footnote ⁹ initially arose as a classic defamation and right of publicity case. Someone on the social media platform Facebook posted a photograph of a famous politician along with personal insults. Importantly, the victim in this case not only requested the removal of the illegal content from the platform but also asked Meta to take adequate measures to ensure that the same or similar content would not be posted again in the future. Once again, the question of prior control of content through filtering, or preventive censorship, arose. The Court had to evaluate the legality of such an injunction.

The judges accepted that such an injunction may be legal, provided that it is not interpreted as imposing on the host provider an obligation to generally monitor the information it stores or to actively seek facts or circumstances indicating illegal activity. In other words, the provider would only have an obligation to make reasonable efforts to block the same content.

This case is particularly important in the context of this chapter as it explicitly addresses the issue of the global or regional reach of EU internet law. Regarding the question of the worldwide or regional impact of such an injunction, the ECJ concluded that the E-Commerce Directive does not impose any limitation, including a territorial one, on the scope of measures that Member States are entitled to adopt in accordance with the directive.Footnote ¹⁰ Therefore, there is no reason to limit the territorial impact of such an injunction.

12.2.3 The Digital Service Act (DSA)

The substantial impact of the DSA is well documented,Footnote ¹¹ as research already pointed out before the regulation was adopted.Footnote ¹² The regulation substantially built on the acquis of the E-Commerce Directive, preferring a system of self-moderation to state-censorship. Under this system, the primary responsibility for notifying and reporting content that allegedly violates the terms and conditions of the service lies with the users of the online service. The platform, in turn, has a duty to respond and intervene as necessary.

At the forefront of this attention is once again the adoption of content-filtering technology. For very large online platforms (VLOPs) and search engines, defined by a set of strict economic conditions, the DSA takes a significant step away from the classic safe harbour approach and imposes a duty to mitigate the risks of illegal content proliferation.

Article 34 introduces a risk assessment, designed to tailor the VLOP’s obligations to the specific dangers associated with their platforms. Article 35 builds upon this risk assessment by requiring VLOPs to implement ‘reasonable, proportionate, and effective mitigation measures’ to address these risks.

Officially, the DSA is presented as an act aimed at protecting digital sovereignty. However, the global reach of such a transformative regulation is evident and not hidden. For example, the EU Commission’s website states that one of the DSA’s two goals is ‘to establish a level playing field to foster innovation, growth, and competitiveness, both in the European Single Market and globally’.Footnote ¹³ Consequently, it has already been observed that this practical function of the DSA as the global regulator of the internet ‘will likely provoke a clash in free expression standards between the U.S. and the EU’.Footnote ¹⁴

12.2.4 Online Hate Speech

There is a strong international legal background concerning the prevention of hate speech proliferation online. The most fundamental milestone at the international level is the first Additional Protocol to the Convention of Europe, commonly known as the Budapest Convention, on cybercrime.Footnote ¹⁵ Nevertheless, hate speech does constitute one of the most sensitive issues between the US and EU from the perspective of the delimitation of the boundaries of freedom of expression. It should be noted that the US, while participating in the drafting of the text, chose not to sign the first Additional Protocol covering the criminalisation of acts of a racist and xenophobic nature committed through computer systems.Footnote ¹⁶ In parallel, in the EU, the EU legislator decided to integrate this criminalisation of public incitement to violence or hatred into EU law. Council Framework Decision 2008/913 states:Footnote ¹⁷ ‘Each Member State shall take the measures necessary to ensure that the following intentional conduct is punishable: (a) publicly inciting to violence or hatred directed against a group of persons or a member of such a group defined by reference to race, colour, religion, descent, or national or ethnic origin’.Footnote ¹⁸

In contrast, in the US, hate speech is protected by the First Amendment. For example, in a case related to a trademark allegedly using a racial slur, the Supreme Court determined that the registration office was not entitled to refuse its registration. The Court held that ‘Speech that demeans on the basis of race, ethnicity, gender, religion, age, disability, or any other similar ground is hateful; but the proudest boast of our free speech jurisprudence is that we protect the freedom to express the thought that we hate’.Footnote ¹⁹

The EU legal framework is not limited to the criminalisation of hate speech. Through soft law, specifically the EU Code of Conduct against online hate speech, the EU legislator aimed to engage major internet actors in the fight against hate speech. Furthermore, with the support of new powers provided by the DSA, the Commission may enforce this code of conduct and monitor its application. The DSA specifically mentions ‘illegal hate speech’ as one area where the mitigation duty of VLOPs must differ from general moderation policies, particularly in terms of speed and effectiveness (Article 35(1)(c) of the DSA refers to expeditious removal or disabling access to hate speech, and adds that dedicated resources for content moderation must be adopted). These efforts aim to compel internet service providers to structurally modify their services and integrate hate speech moderation in their design.

12.2.5 The Right to Be Forgotten in the Google Spain Jurisprudence

The circumstances and findings of the landmark Google Spain decision are well known.Footnote ²⁰ The ECJ determined that freedom of expression, specifically the right to be informed, should yield to the data subject’s right to ‘oblivion’, and in certain circumstances, the data subject is entitled to request that a search engine remove online content that reveals their personal data.

It should be immediately noted that this limitation on freedom of expression is not absolute. First, the right pertains not to the existence of the content itself online but only to it being referenced by search engines. Second, in light of the principle of proportionality, it must be demonstrated that there is an effective legitimate interest justifying the request for de-referencing and that, conversely, the information is not of sufficient public interest for the right to be informed to prevail.Footnote ²¹

Therefore, the right to be forgotten is aptly named, as time itself is not the sole factor in assessing the various legitimate interests. For instance, the accuracy of the information plays a more crucial role. In cases involving fake news, for example, search engines may be compelled to comply with de-referencing requests, as evidenced in the case of Google (Déréférencement d’un contenu prétendument inexact).Footnote ²² Nevertheless, the passage of time remains significant, assuming that the public interest in current information decreases over time.

Once again, the practical importance of this right raises questions about its reach. Should the right influence the behaviour of search engines worldwide, or should it only have regional applicability? This issue was addressed in the Google v. CNIL case,Footnote ²³ where, this time, the theory of global reach was rejected by the Court. Specifically, the ECJ concluded that an:

12.2.6 The Fight against Child Pornography

It might have been expected that a consensus would be reached on combating the highly sensitive issue of child pornography. Indeed, under the framework of the Budapest Convention on cybercrime, both the US and the EU agree on repressing this form of communication. However, an issue has been left to the appreciation of the convention’s signatories, illustrating the extent of the conceptual divergences between Americans and Europeans regarding the topic of freedom of expression.

Indeed, in EU Directive 2011/93/EU on combating the sexual abuse and sexual exploitation of children and child pornography,Footnote ²⁴ child pornography is defined in four alternative ways, aligning with the definition provided by the Budapest Convention on cybercrime. More interestingly, the fourth alternative definition considers as child pornography ‘realistic images of a child engaged in sexually explicit conduct or realistic images of the sexual organs of a child, for primarily sexual purposes’. This means that even in the absence of real abuse, the mere drawing or computer-imaging of a situation of child abuse is considered child pornography. This definition therefore encompasses various modern techniques such as deepfakes (replacing a face in a video with another),Footnote ²⁵ or AI-generated pictures and videos.Footnote ²⁶ By contrast, in the US, the case of Ashcroft v. Free Speech Coalition limits the range of the definition of child pornography to real content only,Footnote ²⁷ considering that in the case of imaginary content (even realistic) freedom of speech should prevail. This stance is generally considered incompatible with the Budapest Convention, which acknowledges virtual child pornography and calls for its criminalisation.Footnote ²⁸ However, on the aftermath of the decision in Ashcroft, the American Congress added what could be seen as a very restrictive interpretation of the Convention, amending the law and limiting the ban to any visual depiction ‘that is, or is indistinguishable from, that of a minor engaging in sexually explicit conduct’.Footnote ²⁹

Moreover, the EU has engaged in discussions on a more active implication of online platforms in fighting child pornography. More specifically, the EU Commission has proposed a new regulation in 2022 to Prevent and Combat Child Sexual Abuse (CSA).Footnote ³⁰ The text has created controversy and the opposition of the EU Parliament and the Council, as it aims to impose advanced moderation and detection duties to online service providers. The complimentary impact assessment produced by the European Parliament’s Committee on Civil Liberties, Justice and Home Affairs was, in particular, extremely critical of the proposal.Footnote ³¹ It details how the new legislation would infringe the prohibition of general monitoring and the prohibition of general data retention principles that are enshrined in EU law and developed in the ECJ’s case law. While a detailed analysis of this proposal and of the subsequent criticism would exceed the limits of this present research, it should be mentioned that the CSA regulation, if adopted, would have substantial consequences worldwide, as it would force online service providers to rethink how and when encrypted end-to-end communication services are provided to users.

12.3.1 Freedom of Expression versus Freedom of Speech

While there is consensus that freedom of expression is one of the fundamental pillars of democracy, this unanimity conceals a significant divergence of viewpoints regarding its definition.

In the US, freedom of speech is defined with reference to the famous theory of the marketplace of ideas. In a dissenting opinion in the landmark case of Abrams v. United States,Footnote ³² Justice Oliver Wendell Holmes argued that the First Amendment protects the right to dissent in response to the government’s viewpoints and objectives. The theory of the marketplace of ideas posits that truth and by extension the common good will naturally emerge from the free exchange of ideas. In other words, all ideas should be freely expressed because they all contribute to the formation of a well-informed public opinion. In 1969, the US Supreme Court’s decision in Brandenburg v. Ohio solidified this theory as the dominant public policy in US free speech law,Footnote ³³ explaining that the government cannot penalise inflammatory speech unless that speech is ‘directed at inciting or producing imminent lawless action and is likely to incite or produce such action’.

Nevertheless, multiple authors have warned against an overly literal interpretation of the theory of the marketplace of ideas. It has been pointed out that the classical interpretation depends on impossible assumptions for its coherence,Footnote ³⁴ and Gordon in 1997 demonstrated that the theory, taken absolutely, would only create a situation where the views of the most powerful and/or the most numerous prevail.Footnote ³⁵ In consequence, the theory of the marketplace of ideas has evolved over time. For instance, in the landmark Edwards v. Aguillard (1987) case, the US Supreme Court considered that a law stating that if one theory is exposed at school, the other theory must also be presented (referring to the principle of natural evolution, as opposed to the dogma of creationism), is unconstitutional.Footnote ³⁶

In Europe, on the contrary, the indelible scars of the Holocaust have led to the adoption of a rather different stance on freedom of expression. In this context, a free marketplace of ideas is considered a utopia. Under the prism, implicitly, of the theory of memes,Footnote ³⁷ society is seen as an ecosystem where some ideas act as viruses, contaminating people’s minds and jeopardising democracy. Therefore, freedom of expression, as enshrined in Article 10 of the Convention on Human Rights and Article 11 of the Charter of Fundamental Rights, is not perceived as absolute, as illustrated in the famous ‘little red schoolbook’ case.Footnote ³⁸ While restrictions on freedom of expression are accepted, they must pass a three-part test, as recast in cases such as Mouvement Raëlien Suisse v. Switzerland,Footnote ³⁹ and Animal Defenders International v. the United Kingdom;Footnote ⁴⁰ the limitation must be provided by law, pursue a legitimate aim, and be necessary in a democratic society. This last condition has been further explained by case law, with emphasis on the notion of the ‘pressing need of society’. In other words, for a restriction to freedom of expression to be legitimate in the European perspective, it is not enough for the restriction to serve a public interest. The restriction must be unavoidable in a democratic society. All the restrictions analysed above (copyright law, child pornography, protection of reputation, data protection, hate speech) have been subjected to the three-part test.

It is commonly accepted that the American philosophy of freedom of speech is much broader than the European one, but it is essential to temper this notion, as in some respects, the American concept is more restrictive than the European one. First, the American human rights framework has primarily been constructed as a defence mechanism against state intrusion, under the ‘state action doctrine’,Footnote ⁴¹ resulting in the impossibility of horizontal application. Meanwhile, in European traditions, some forms of horizontal application are permitted.Footnote ⁴² Second, American freedom of speech is largely conceived from an individualistic standpoint, whereas the European perspective could be described as multidimensional, encompassing not only the freedom to express ideas but also the protection of pluralism,Footnote ⁴³ and the right to be informed.Footnote ⁴⁴

In conclusion, the American and European approaches to freedom of expression converge in terms of their core, which is the right to provoke. However, the various social functions of freedom, its role in society, and, most importantly, its limits are subjects of significant divergence between the birthplace of the internet and the EU.

12.3.2 Case Study: The Situation of X (formerly Twitter)

The recent policy changes of X (formerly Twitter) following its acquisition by Elon Musk should be used as a case study to illustrate the potential conflicts that can arise from this divergence. Indeed, the new president of the famous social media platform, who is a self-declared free speech absolutist,Footnote ⁴⁵ has chosen to implement an orthodox American freedom of speech’s perspective on X’s functioning and moderation policies.

Consequently, the company lost around half of its workforce in 2022 (among them a substantial percentage were employees who focused on ethics and moderation) and then started to reinstall contentious accounts that had been suspended. Unsurprisingly, the platform almost immediately saw a rise in hate speech in its content.Footnote ⁴⁶

However, according to Twitter’s advertising audience data from July 2022, the social network currently has 30.6 million users in Western Europe, and 27.2 million in Northern Europe.Footnote ⁴⁷ Therefore, it is not surprising that, when Elon Musk tweeted ‘the bird is freed’ (after acquiring the company), the Internal Market Commissioner Thierry Breton replied ‘[i]n Europe, the bird will fly by our rules’.Footnote ⁴⁸ Furthermore, in 2023, the European Commission Vice President focused on X in a press statement, highlighting the misfunctioning moderation system of the platform.Footnote ⁴⁹

Indeed, data shows that X is having the ‘largest ratio of mis/disinformation posts’ among platforms that submitted reports to the EU. In the context of information wars and while the existence of ‘fake news factories’ have been demonstrated,Footnote ⁵⁰ the reactions of the EU and the US diverge again, as they are based on their definition of freedom of expression. The EU pushes for stricter control of posted content; for instance, by putting forward the model of the ‘trust flagger’ in the DSA, organisms charged with the mission to accompany and guide the moderation teams of the online platforms.

By contrast, in the US, the case United States v. Alvarez,Footnote ⁵¹ where the Supreme Court ruled that the crime of falsely claiming military honour is unconstitutional, clearly stated that even disinformation is by presumption protected by free speech. Only under strict scrutiny may a content-based restriction of freedom of speech be accepted by exception in the US.Footnote ⁵²

12.4.1 The Danger of a Fragmented Internet

Based on the observations in Section 12.2, it can be concluded that the EU legislator, in conjunction with jurisprudence, does not hold an absolute position on the question of whether EU internet law should have regional or global reach. While the CNIL decision accepted that the right to be forgotten could only have regional reach, the Eva Glawischnig case, on the contrary, applied global reach. The situation is even more complicated concerning the mitigation duty stemming from the DSA and the Article 17 filter, where no indication exists regarding the reach of the legal framework. Generally speaking, any legislation that would allow for regional reach alone would be structurally flawed, as users could easily circumvent such regulations by using a virtual private network (VPN). Nevertheless, the National Security Agency revelations in the past have potentially damaged trust on internet governance in Europe, and ever since there has existed a deep current towards global reach in EU regulation of the internet, leading to a fragmentation of the internet.Footnote ⁵³

Furthermore, online law enforcement is also a matter of the protection of democratic values for EU institutions. Characteristically, in the 2023 European Declaration on Digital Rights and Principles for the Digital Decade,Footnote ⁵⁴ the EU institutions declared with force that ‘the digital transformation should not entail the regression of rights. What is illegal offline, is illegal online.’Footnote ⁵⁵ Nonetheless, the global character of the digital environment pushes the boundaries of private international law. For instance, the GDPR explicitly considers it applies from the moment that EU data subjects are specifically targeted by the processing, even if neither the data controller nor the processing is in the EU. In any case, it has to be accepted that in any matter that concerns criminal law (such as hate speech), the EU is also entitled to defend the enforcement of its legal framework in the digital environment.

From a practical standpoint, it is doubtful that major companies would invest substantial resources in compliance with legislation that applies regionally or risk non-compliance with various European obligations, only to find out that European courts view them as obliged to comply because European citizens are indirectly affected by their activities. This phenomenon can be observed in the functioning over time of some of the most popular generated language artificial intelligence models, such as ChatGPT. Various white papers show that the company has taken a precautionary approach regarding mitigation of the risks of generating ‘illicit content’, including safeguarding anti-racist and anti-hate speech content.Footnote ⁵⁶

As a consequence, American proponents of the internet face three options: comply with the EU legislator’s expectations worldwide, ignore them and limit their market, potentially losing one of the largest digital markets in the world, or offer differentiated services depending on the user’s location. The third option may seem appealing at first: IT companies would not lose any market or antagonise any legislator. The simplest solution would be to distinguish the service based on user internet protocol (IP) address, or even implement geo-blocking. For instance, some American newspapers automatically block access to viewers with European IP address, fearing the adoption of the GDPR.Footnote ⁵⁷ However, this ‘soft division’ of the internet provides more legal uncertainties. Indeed, as the use of VPNs constantly increases in the population, the general awareness of the ‘geo-unblocking’ capabilities of such networks also expands,Footnote ⁵⁸ to the extent that in a foreseeable future any geo-blocking attempt will be deemed inefficient on the prism of the principle of proportionality.

Alternatively, there is a risk of a ‘hard division’ or fragmentation of the internet. China for instance, following its own agenda of absolute national control of communications, has developed native alternative internet technologies that could splinter the global internet’s shared and ubiquitous architecture.Footnote ⁵⁹

It has been argued that the territorialisation of cyberspace does not necessarily entail risking the fragmentation of the internet.Footnote ⁶⁰ However, the arguments proposed are unconvincing because it is challenging to ignore how, in a globalised context, efforts to protect digital sovereignty can be seen as a form of digital imperialism.

A high-stakes game is currently unfolding – while Europe would not benefit from the GAFAMs leaving the internal market, the companies also have no incentive to abandon it. In this context, Meta’s behaviour, first threatening to leave the market after the US-EU Safe Harbor and Privacy Shield agreements were struck down by the ECJ,Footnote ⁶¹ and then launching its Twitter-like social network Threads in the EU, should be seen as diplomatic attempts to push back against the EU legislator’s imperialist view of the digital landscape.

12.4.2 The Need for an International Structure of Conciliation

Several international organisations have shown interest in resolving the challenge of enforcing laws in the digital realm without infringing upon online freedom of expression. The Council of Europe, in particular, has a long history of issuing guidelines on this topic.Footnote ⁶² In general, the question of global internet governance has been discussed as long as the internet has existed. It could be proposed that the United Nations is a natural choice for this debate, especially through its action via the Internet Governance Forum (IGF),Footnote ⁶³ which embraces the ‘multistakeholder model’ (with all its flaws).Footnote ⁶⁴ However, the actions of the IGF and the limitations of its mandate suggest that it may not be an effective platform for discussing the boundaries of online freedom of expression.

In this context, while the choice of the WTO may seem unusual at first considering it is primarily an economic organisation, there are arguments in favour of this option. First, the interstate Dispute Settlement Understanding is a highly efficient mechanism, and it is believed that enforcement through the WTO has been easier than through other international instruments.Footnote ⁶⁵ Second, the divergent definitions of online freedom of expression impose significant economic costs on IT companies, which can be viewed as an economic factor. It has thus been argued that the WTO has a role to play in limiting China’s online censorship policies.Footnote ⁶⁶ Third, while differences in perspectives on freedom of expression are deeply rooted in constitutional systems, the WTO is seemingly independent of specific legal traditions. While the search for a synthesis between EU and US approaches to online freedom of expression involves complex reasoning and considerable effort, the use of a comparative approach privileged by the organisation would provide some insights on the issue. Finally, the human rights approach is not foreign to the WTO, and how the WTO is involved in resolving disputes related to freedom of expression has already been analysed,Footnote ⁶⁷ using the concept of ‘public morals’ as a legitimate limitation to freedom of expression.Footnote ⁶⁸

12.4.3 Through the Protection of Digital Public Spaces

Finding a structure capable of resolving international disputes over online freedom of expression is not enough; it should also be recognised that the existing definitions, both American and European, are insufficient to fully protect online freedom of expression. In other words, online freedom of expression requires not only a structure but also substantial legal adjustments. One of the fundamental aspects of the protection of online freedom of expression today is its heavy reliance on free access to social media platforms. However, these platforms are owned by private companies, which raises new challenges for the freedom of expression. Some authors then refer to the ‘platformisation of the public sphere’.Footnote ⁶⁹ Digital public spaces are governed by algorithms of presentation and management of the information flow. The huge potential of these algorithms in influencing public opinion has been demonstrated. The 2023 Declaration on Digital Rights and Principles expressly mentions this issue, stating that:

Furthermore, it is also argued that the digitalisation of the public space entails ‘antagonistic and networked-individualistic flows of populist communication’,Footnote ⁷¹ and effectively, the bubble effect of social media is nowadays well documented.Footnote ⁷²

One of these issues is the need for the horizontal application of freedom of expression. Here, ‘horizontal effect’ refers to the ability of end users to initiate private litigation against an online platform for violating their human rights, specifically by filing a complaint when their freedom of expression has been infringed owing to abusive moderation. In the US, no margin of appreciation is recognised for the horizontal effect of freedom of expression. For example, in the District Court in California FAN v. Facebook case,Footnote ⁷³ it was ruled that private internet service providers such as Facebook are entitled to shut down users’ accounts, as social media does not operate as a public forum.

By contrast, the EU tradition allows for greater application regarding the horizontal effect of freedom of expression (see for instance the Viking Line judgement,Footnote ⁷⁴ where the ECJ adopted the German theory of the indirect horizontal effect of human rights). Furthermore, the landmark Egenberger and Bauer cases directly acknowledge a direct horizontal effect to some provisions of the Charter (the principle of non-discrimination and effective judicial protection).Footnote ⁷⁵ These two cases have substantial importance for EU constitutional law,Footnote ⁷⁶ and this reasoning could be seen as opening the door for further recognition of the horizontal effect. However, it is uncertain, in the absence of relevant secondary EU legislation, whether the reasoning employed in these two cases can be used by extension for freedom of expression.

Therefore, in the US without any doubt, and in the EU most probably, the enforcement of online freedom of expression against private actors, even those managing very large social platforms and acting as de facto gatekeepers of public spaces, remains uncertain. This has very practical consequences: How is it possible to prohibit a platform from banning an account or imposing a ‘shadow ban’ on a communication (through algorithmic moderation to restrict the range of a publication), thus manipulating opinion and enforcing private censorship, without having any control? Shadow banning, as a conscious form of moderation, might be a myth.Footnote ⁷⁷ Nonetheless, the automatic consequence of algorithmic information presentation is that certain content, based on predefined criteria, may experience reduced visibility. In the EU, the DSA constitutes a first step towards regulation of this practice.Footnote ⁷⁸

Consequently, the protection of online freedom of expression currently relies on an indirect and imperfect legal framework, which combines contract law with consumer protection. For example, in Germany, in the Facebook Terms of Service Case,Footnote ⁷⁹ it was determined that Facebook’s terms of service on deleting user posts and blocking accounts for violations of its Community Standards were invalid. Similarly, in France, in the ‘nombril du monde’ case,Footnote ⁸⁰ the court found that by moderating the posting of a painting without explanations, Facebook had violated the general duty of good faith in the execution of the contract, as mandated by the Civil Code.

This situation is far from satisfactory. The philosophy and methodology of contract law cannot be seen as an ideal method to guarantee freedom of expression. It is time for a re-evaluation of online freedom of expression that takes into consideration the practical reality of some social media platforms as digital public spaces. As a result, no obstacles should be placed against the enforcement of freedom of expression in these digital public spaces.