Francesca Cassano (Università degli Studi di Milano)

1. Introduction

On January 7^th, 2025, Mark Zuckerberg, CEO of Meta (hereafter also referred to as the “Company” or the “Platform”), announced several changes to the content moderation policies governing its social media platforms, namely Facebook, Instagram and Threads. These reforms addressed two key aspects of the Company’s content moderation practices: the fact-checking program and the so-called Hate Speech Community Standards. Regarding the former, Meta decided to eliminate, starting in the US, the fact-checking program and replace it with a community notes system, similar to the one already implemented on platform X. As for the latter, the Hate Speech Community Standards were simplified by removing a series of restrictions on topics deemed to be, allegedly, of public interest. Specifically, the Company’s decisions aim, at least ostensibly, at addressing and preventing an allegedly excessive censorship that has characterized the Platform’s content moderation to date, as well as to avoid silencing diverse opinions and ideas on «mainstream» issues. However, as it will be argued, it seems that the declared intention of the Company at better «promoting free speech» is just the latest example of a social media company aligning to the policies of the new US administration.

As a matter of fact, regardless of the Company’s real purposes, the amendments implemented raise several questions about their compliance with human rights law standards and their potential negative impacts towards groups that are already primary targets of hate speech on social media platforms. Certainly, even before these reforms, Meta/Facebook was not known for its strong commitment to human rights, and its role in alleged human rights violations has been widely criticized. Several tragic episodes — such as the genocide of the Rohingya in Myanmar, the anti-Muslim riots in Sri Lanka, and the ethnic cleansing of Tigrayan civilians in Ethiopia — have unfortunately demonstrated the Platform’s involvement in human rights abuses and in the commission of offline violence against vulnerable and marginalized groups. However, despite its failure to properly address these episodes and prevent them, it appeared that the Company was at least attempting to make some progress, albeit without fully addressing the issues. Initiatives such as the establishment of the Oversight Board in 2020 and the adoption of a Corporate Human Rights Policy in 2021 suggested an effort, albeit limited, to improve the Company’s human rights commitment.

Nevertheless, rather than a step forward, this latest move represents a deliberate regression that risks entrenching harmful practices under the guise of freedom-oriented reforms. In this regard, specific concerns have been raised regarding Zuckerberg’s statement about the need to «get rid of a bunch of restrictions on topics such as immigration and gender that are just out of touch with mainstream discourse» (see announcement of January 7^th). Indeed, following Meta’s announcement, the UN High Commissioner for Human Rights criticized the Platform’s decisions, emphasizing that «allowing hate speech and harmful content online has real-world consequences» and that «regulating this content is not censorship».   Against this critical background, the following paragraphs will first provide a brief overview of the existing human rights law standards on hate speech applicable to online platforms. The focus will then shift to assessing the compliance of the new Meta’s Community Standards on Hateful Conduct with this framework, as well as the concrete threats they may pose to vulnerable groups and marginalized communities.

2. Human rights law standards regulating hate speech and their application to online platforms

When considering the regulation of hate speech within the human rights law framework, there are several important, albeit few, provisions found in both universal and regional human rights law treaties. It should be preliminarly emphasized that none of the following provisions explicitly mention the term “hate speech”, that to date does not have a universally accepted definition, nor do they directly bind Meta or other online platforms, as they are binding only towards their member States. However, analysing these provisions is crucial for understanding the human rights law standards on hate speech, which online platforms, such as Meta, should consider when developing and enforcing their own content moderation policies.

Regarding the universal framework, key provisions are outlined in the International Covenant on Civil and Political Rights (“ICCPR”), particularly in Art. 19, par. 3 and in Art. 20, par. 2, as well as in the International Convention on the Elimination of All Forms of Racial Discrimination (“ICERD”) in Art. 4. In particular, Art. 19 par. 3 of the ICCPR provides the conditions under which States parties may restrict the right to freedom of expression (protected in par. 1 and 2 of the same article) in order to protect national security, public order or the rights and reputations of others. On the other hand, Art. 20 par. 2 of the ICCPR and Art. 4 of the ICERD go further by imposing specific positive obligations on States parties to prohibit and criminalise certain forms of incitement to hatred (for a detailed analysis of this framework see De Sena and Castellaneta). At the regional level, particular attention should be given to the framework of the European Convention on Human Rights (“ECHR”). While the ECHR does not explicitly include a provision equivalent to Art. 20 par. 2 of the ICCPR or Art. 4 of the ICERD — both of which prohibit and criminalise incitement to hatred, as does Art. 13, par. 5 of the American Convention on Human Rights — the European Court of Human Rights (“ECtHR”) has, nonetheless, developed the most extensive jurisprudence on hate speech to date (see here). In particular, the ECtHR, in applying the provision that prohibits the abuse of rights under Art. 17 of the ECHR, has effectively excluded from the scope of Art. 10 — which guarantees the right to freedom of expression — all forms of hate speech that are manifestly contrary to the principles and values upheld by the Convention (see Castellaneta). The ECtHR jurisprudence has also played a crucial role in shaping the responsibility of online platforms in regulating harmful content. A landmark case in this regard is Delfi AS v. Estonia (2015), in which the Court held that online platforms could be held liable for hosting offensive comments, even if the platform itself did not directly post the content and that they must also take proactive measures to prevent the dissemination of hate speech and other harmful material (see also in this regard Sanchez v. France and for an analysis of this latest decision see Castellaneta).

Turning now to the standards applicable to online platforms, as it has been explicitly confirmed by the 2018 report of the UN Special Rapporteur on the promotion and protection of the right to freedom of opinion and expression, social media companies’ content moderation should align with international human rights law. This framework derives primarily from the 2011 UN Guiding Principles on Business and Human Rights (“UNGPs”), which call on all businesses — regardless of their type or size — to respect human rights, implement a human rights due diligence mechanism (“HRDD”), and address adverse human rights impacts, including by ensuring access to effective remedies for the victims of their violations. Beyond the UN framework, another key instrument that applies both to multinational and domestic enterprises are the OECD Guidelines for Multinational Enterprises on Responsible Business Conduct (“OECD Guidelines”). A particularly notable feature of this instrument is the National Contact Point system (“NCP”), which serve as non-judicial dispute resolution mechanisms aimed at overseeing the implementation and effectiveness of the OECD Guidelines. To date, NCPs have rarely handled cases involving online platforms. However, a complaint is currently pending before the United States NCP against Facebook/Meta, after initially being filed before the Ireland NCP. The complaint calls on the Platform to provide a remedy to the Rohingya people for its role in hosting content that incited violence and hatred against the minority and that contributed to the 2017 “clearance operation” carried out by the Burmese army.

There have also been more targeted efforts of the international community to engage online platforms in countering hate speech and improving their transparency and accountability. For instance, a recent initiative was led by UNESCO, which in 2023 published the Guidelines for the Governance of Digital Platforms. According to these guidelines, online platforms should comply with several key principles such as assessing their human rights impact, including on gender and minority issues (par. 85–90); they also specifically urge digital platforms to «put in place sufficient special protections for women and girls, users from groups in situations of vulnerability and marginalization» (par. 130). Furthermore, at the regional level, highly significant is the Recommendation on combating hate speech, adopted by the Committee of Ministers of the Council of Europe in 2022. One of the most compelling aspects of this Recommendation is the suggestion for all stakeholders, including online platforms, to calibrate their response to hate speech by considering not only the usual contextual factors such as the mode of dissemination and the size of the audience, but also the specific harms and impacts of the phenomenon, as well as the specific characteristics of the targeted groups (par. 3 and 4, and for recommendations regarding Internet intermediaries, see from par. 30 to 37).

However, as already mentioned, all the aforementioned instruments are non-binding for online platforms, as International Law has traditionally been conceived as binding primarily on sovereign States. This means that non-state actors, such as transnational companies, do not have corresponding international legal obligations. For this reason, it is also crucial to briefly consider the significant developments within the European Union law framework in its efforts to directly regulate businesses, and more specifically, online platforms. Accordingly, recent initiatives such as the Corporate Sustainability Due Diligence Directive (“CSDDD”) and the Digital Services Act (“DSA”) aim at imposing stricter obligations on companies, enhancing accountability and ensuring greater alignment with human rights standards also in the digital sphere. The CSDDD also applies to non-EU companies with a turnover exceeding 450 million euro in the EU in the last financial year, requiring them to integrate a HRDD that help to prevent and mitigate their adverse human rights impacts (for a detailed analysis see Bonfanti and Greco). The DSA holds Very Large Online Platforms (“VLOPs”) — those with an average number of monthly active service recipients in the EU equal to or exceeding 45 million — accountable by requiring them to assess and mitigate systemic risks related to illegal content, including hate speech (Art. 34). If a risk is identified, Art. 35 par. 1 lett. c allows platforms to take measures such as content removal, but only upon prior notification by a user or trusted flagger (Artt. 16, 22) (see in this regard Ruotolo). Consequently, for major online platforms falling within their scope, such as Meta, this regulatory framework may directly affect their online moderation of hate speech. Moreover, it appear also to be the reason why the new Meta’s community note system has been implemented only in the US so far and also the reason behind Zuckerberg’s declaration on his intention to «work with President Trump to push back governments around the world, going after American companies and pushing to censor more» (see announcement of January 7^th).

Ultimately, while the said human rights law framework sets important principles on hate speech regarding the content moderation processes of online platforms’, its limited enforceability highlights the need for more robust and binding regulations, as seen in the EU’s emerging legal landscape. Overall, these efforts contribute to countering hate speech and addressing online platforms’ responsibility, though they are not without flaws, particularly in terms of effectiveness, clarity, and the absence of a coherent and unified approach in handling the phenomenon.

3. Assessing the compliance of the new Meta’s “Hateful Conduct Community Standards” with the human rights law framework on hate speech

In light of the aforementioned legal framework, it is now paramount to highlight the main issues arising from Meta’s new Community Standards on Hateful Conduct. It should be preliminarily noted that to assess the Company’s reforms, a detailed description of all amendments in the new Community Standards is available on Meta’s Transparency Center website.

A first important consideration is the decision to change the policy name from “Hate Speech Community Standards” to “Hateful Conduct Community Standards”. Although this shift may seem merely terminological at first glance, when considered within the current political and cultural context, it reveals a potential redirection and narrowing of the policy’s scope of application. Specifically, the new name suggests that the policy may now only target expressions that are directly linked to material acts of violence. Indeed, this approach seems to mirror a US centric interpretation of free speech, which tends to restrict expressions only when they incite or are closely connected to violent actions (the said principle was stated by the US Supreme Court in 1969 in the Brandenburg v. Ohio case). This, in turn, risks undermining the broader protections against hate speech set out in international human rights law, which also calls for the prohibition and/or criminalisation of speech that fosters discrimination and hostility, not only violence (see Art. 20 par. 2 ICCPR and Art. 4 ICERD). Moreover, even if the definition of “hateful conduct” contained in the new Standards does not differ from the previous one on “hate speech” – namely a direct attack against people on the basis of specific protected characteristics – the connection between the spread of hate speech and the promotion of an environment of intimidation, exclusion and of potential offline violence has been removed. This omission itself shows an intention of downplaying the broader societal harms of online hate speech, thus weakening the Platform’s responsibility to mitigate its potential real-world consequences.

Moreover, within the new Standards, the Platform then classifies the severity of “hateful conduct” into two tiers: the first tier identifies the most severe forms of content that dehumanize or incite serious harm or violence towards individuals or groups, while the second tier targets content that attacks individuals or groups based on their protected characteristics. In this regard, what differs with the previous Standards and what emerges as more concerning isn’t the number of tiers – that were already modified in December 2023 passing from three to only two levels of severity – but the removal within them of a series of protections towards groups that are historically oppressed and marginalized. For instance, among the most alarming policy changes, the Platform now permits users to advocate for exclusion or use insulting language when discussing “political” or “religious” topics, such as transgender rights, immigration, or homosexuality.  In addition, while Meta’s policy still prohibits users from posting content that targets individuals or groups based on their protected characteristics or immigration status with dehumanizing speech — such as comparisons to animals, pathogens, or sub-human life forms — the recent changes imply that it may now be permissible to compare women to household objects as well as allowing the use of dehumanising pronouns like “it” for transgender and non-binary individuals. Furthermore, the Platform now permits to its users to describe LGBTQ+ people as mentally ill or abnormal and to call for their exclusion from professions, public spaces, and society based on their sexual orientation and gender identity. Such changes not only pose a concrete threat to the inclusions of individuals and of members of protected groups, but they also clearly fail to align to the above-mentioned human rights law framework, which explicitly suggests online platforms to consider the human rights impacts of their policies on vulnerable individuals and communities (see, for instance, Principle No. 18 UGPDs, as well as par. 89 and 130 of the UNESCO Guidelines).

It is unfortunately evident that Meta’s new approach, framed as a commitment to better guarantee the right of freedom of expression, not only fails to align with the aforementioned business responsibility to respect human rights, but also undermines the very principle it professes to protect. Indeed, it is well established that hate speech — both online and offline — not only causes physical, psychological, and economic harm to its targets, but also compromises their sense of security and may silence them out of fear of potential repercussions (see Matsuda et al.). Moreover, hate speech, especially if widespread and uncontrolled such as on social media platforms, may represent a direct threat towards political and social values and it may undermine principles upon which constitutional democracies are founded (see Waldron). It should also be noted that the new Standards constitute a clear retreat from the Human Rights Policy that the Company itself adopted in 2021. Among its human rights commitments, Meta explicitly pledged to «pay particular attention to the rights and needs of users from groups or populations that may be at heightened risk of becoming vulnerable or marginalized» (see commitment no. 2). However, with the 2025 hateful conduct policy the Platform clearly jeopardises these earlier human rights commitments by ignoring and removing crucial protections for vulnerable groups. In other words, the Platform does not appear to have adopted an effective HRDD process. This, unfortunately, highlights the limitations of the aforementioned standards, as their implementation remains at Meta’s discretion, with no oversight or accountability ensuring their actual enforcement.

In this pressing scenario, many questions should arise about the role that the Meta Oversight Board will play in supervising the Platform’s content moderation processes, particularly regarding whether it might have the authority to mitigate the potential negative impacts of the 2025 Hateful Conduct Community Standards. As previously mentioned, the Oversight Board was established by the Company in 2020 after facing immense pressure from users, governments, and civil society to act more transparently and with accountability. Recently, this instrument has been defined as a form of “transnational hybrid adjudication” in the sense that it represents an independent dispute settlement body that resolve cross-border legal cases (see Gulati). In particular, according to the introduction of the Oversight Board Charter, its purpose is to «protect free expression by making principled, independent decisions about important pieces of content and by issuing policy advisor opinions on Meta’s content policy». In other words, the members of the Oversight Board determine whether the Platform was justified in removing or leaving up controversial content. Moreover, the Oversight Board’s decisions in that matter are binding upon Meta, meaning that the Company is required to implement its ruling on content moderation cases. In this regard, it is also essential to specify that, in adjudicating disputes, the Board does not rely only to the Platform’s standards and values but considers also «human rights norms protecting freedom of expression» (Art. 2 section 2 of the Board’s Charter). Thus, international human rights law is explicitly referenced as a source of applicable law for the Oversight Board. This is further confirmed by the Board’s practice, as it often highlights the Company’s human rights responsibilities under instruments such as the UDHR, the ICCPR and the UNGPs when explaining its decisions (see for example these recent Board’s multiple case decisions here and here).

On the other hand, the establishment of the body has not been without criticism, particularly regarding its overall effectiveness (see inter alia Di Stefano and Wong et. al). For instance, concerns remain about the broader systemic impact of another key function of the Oversight Board, i.e., issuing recommendations on how Meta can improve its content moderation policies. This is because, regarding the implementation of the said recommendations, Meta retains discretion over whether and how to enforce the policy changes suggested by the Board, unlike the aforementioned individual content moderation decisions that are binding upon the Platform. Such discretionary power could limit the Board’s ability to drive meaningful and lasting structural reforms. Additionally, another key concern is the limited number of cases the Board reviews compared to the high volume of requests it receives. Given the vast scale of content moderation challenges on Meta’s platforms, the small fraction of cases the Board adjudicates raises questions about its overall effectiveness in shaping platform-wide policy changes. Nevertheless, it should also be acknowledged that, to date, the Oversight Board’s deliberations and activities appear to have effectively incorporated the existing international human rights law framework on hate speech (see Di Stefano) and it also represents an important — albeit singular — effort to address the critical issues raised by platforms’ content moderation decisions (see Gradoni and also Tiedeke et. al).

To date, there haven’t been many reactions from the Oversight Board after the implementation of the new Standards by the Company. In particular, after Zuckerberg’s announcements, the Oversight Board issued a brief statement on the same day, which focused more on the Platform’s decision to eliminate the fact-checking program than on the new hateful conduct policy. Anyway, on that occasion it was also stated by the Board that it «will be reviewing the implications of the various changes in line with its commitment to freedom of expression and other human rights». Consequently, there is a slight hope that the Board will oversee Meta’s content moderation decisions and at least try to ensure to reconcile the Company’s content moderation new practices with the existing international human rights law framework on hate speech. This will be crucial not only in ensuring accountability but also in safeguarding the rights of vulnerable groups and upholding International Law principles on freedom of expression.

4.  Concluding remarks

This recent policy changes not only undermines Meta’s commitments to human rights but appears also to align with a broader political landscape in the United States that prioritize deregulation and corporate discretion over accountability. A major challenge in promoting a human rights-based approach to content moderation on online platforms is that many of these companies are based in the US, where the near-absolute protection of freedom of expression under the First Amendment has resulted in minimal regulation of hate speech. There is a significant contrast between the approaches adopted by the United States and those within the international human rights law framework in regulating hate speech. This domestic legal framework significantly influences how these platforms define and address harmful content, often limiting their commitment to international human rights law standards.

On the other hand, what Zuckerberg and other online platform CEOs must acknowledge is that the activities of their multinational companies extend far beyond the United States, influencing and impacting users worldwide. Therefore, they should, first and foremost, comply with international legal standards on freedom of expression, which, for instance, require not only the prohibition of advocacy of violence but also of discrimination and hostility (Art. 20 par. 2 of the ICCPR).

In conclusion, instead of implementing a reform that clearly undermines the protection of vulnerable individuals and groups, the Platform could have focused more on improving and reinforcing oversight mechanisms that would also help to genuinely promote the right to freedom of expression by reducing the risks of errors and biases in Meta’s content moderation processes. For instance, if the platform’s true intention had been to protect the freedom of expression of its users, it could have invested in independent human rights impact assessments and stronger internal auditing processes. Additionally, it could have promoted the involvement of civil society organizations, human rights experts, and of affected communities to ensure that its policies and enforcement strategies align with international human rights standards on hate speech (see Hatano). Conversely, the Platform’s recent policy shift appears to be less a principled reassessment of free expression and more a strategic response to shifting political dynamics. In other words, Zuckerberg, after detecting a shift in public attitudes as well as in the US political climate, adjusted Meta’s approach accordingly — disregarding the fundamental rights of vulnerable groups in the process. Ultimately, Meta’s new Community Standards on hateful conduct should have aimed, instead, at both protecting freedom of expression and safeguarding the rights of vulnerable users. Online platforms like Meta must recognize these two goals as cumulative rather than mutually exclusive and ensure that their policies are shaped by fundamental rights rather than shifting political pressures.