Online Hate Speech and International Human Rights Law
- Edmarverson A. Santos
- 4 days ago
- 26 min read
Updated: 3 hours ago
Introduction
Online Hate Speech and International Human Rights Law begins with a basic legal distinction: international law does not impose a general ban on all hateful expression. No universal treaty defines “hate speech” as a single offense, and the term covers forms of communication with very different legal consequences. Some expression must be prohibited. Other speech may be restricted only under strict conditions. Language that is insulting, prejudiced, or degrading may still remain legally protected.
Article 19 of the International Covenant on Civil and Political Rights protects the freedom to seek, receive, and impart information and ideas regardless of frontiers. Article 20(2), however, requires states to prohibit by law advocacy of national, racial, or religious hatred that constitutes incitement to discrimination, hostility, or violence (United Nations, 1966). The two provisions must be read together rather than treated as competing rules.
Other treaties impose distinct obligations. Article 4 of the International Convention on the Elimination of All Forms of Racial Discrimination is more prescriptive in several respects, but its scope is tied to racial discrimination based on race, color, descent, or national or ethnic origin (United Nations, 1965). Direct and public incitement to commit genocide falls within a separate and much narrower category under Article III(c) of the Genocide Convention (United Nations, 1948). These regimes differ in their protected grounds, required intent, anticipated harm, and enforcement consequences.
Expression outside those mandatory prohibitions must ordinarily be assessed under the applicable freedom-of-expression limitations regime. Article 19(3) of the ICCPR permits restrictions only when they are provided by law and necessary for one of the aims recognized in the Covenant. The Human Rights Committee has also emphasized proportionality and warned that restrictions must not jeopardize the right itself (Human Rights Committee, 2011). A government cannot justify censorship merely by describing speech as offensive or harmful.
The online setting changes the factual assessment without creating a separate human rights regime. Digital communication can expand the reach and lifespan of a message through reposting, recommendation systems, coordinated campaigns, and circulation across borders. Meaning and intent may also be harder to assess when expression appears through memes, coded language, irony, quotation, or anonymous accounts. The Rabat Plan of Action offers a non-binding method for evaluating serious incitement by reference to context, speaker, intent, content, extent, and likelihood of harm (OHCHR, 2013).
Major social-media platforms add another layer of regulation. Their community standards may prohibit material that a state could not lawfully suppress, while ranking systems, demonetization, account restrictions, and automated moderation determine what users can see and say. These systems may fail to remove serious abuse, but they may also misclassify satire, counterspeech, reporting, or culturally specific language. International standards on business and human rights can guide corporate conduct, but they do not place platforms under the same treaty obligations as states.
Describing content as “hate speech” does not establish that criminal punishment is lawful, that removal is required, or that a platform’s decision respects human rights. The legal analysis depends on the nature of the expression, the risk it creates, the actor exercising power, the basis for intervention, and the safeguards available to those affected. Without those distinctions, measures intended to protect people from discrimination or violence may enable arbitrary censorship, while an unqualified appeal to freedom of expression may leave serious harm unanswered.
1. Online Hate Speech Under International Human Rights Law
“Hate speech” is widely used by governments, international organizations, civil society groups, researchers, and social-media companies. It does not, however, correspond to a single offense defined across international human rights law. The principal global treaties regulate particular forms of expression through different provisions on incitement, racial discrimination, genocide, personal security, and freedom of expression (McGonagle, 2013; Clooney and Gardoll, 2024).
The label may refer to degrading stereotypes, discriminatory insults, threatening messages, organized harassment, advocacy of hostility, or calls for violence. These forms of expression can overlap, but their legal consequences are not identical. Classification depends on the content of the communication, the speaker’s purpose, the social and political context, the targeted group, the audience, and the harm being encouraged or made more likely.
Digital distribution affects that assessment. A message may be amplified through recommendation systems, repeated by coordinated accounts, or presented through images, symbols, and coded language whose meaning depends on a particular community. Reach and repetition can increase the risk of harm. They do not, by themselves, establish that the expression constitutes unlawful incitement.
1.1 The Absence of a Universal Definition
The absence of a universal definition reflects substantive differences between the forms of conduct commonly placed under the hate-speech label. A credible threat may engage the rights to life and personal security. Persistent, targeted abuse may amount to harassment. Defamation concerns harm to reputation, while discriminatory conduct may interfere with equality in employment, education, housing, or access to services. Each category has its own elements and legal consequences.
Advocacy of hatred is also distinct from incitement. A person may promote hostility toward a group without encouraging listeners to discriminate against, attack, or otherwise harm its members. Article 20(2) of the International Covenant on Civil and Political Rights addresses advocacy of national, racial, or religious hatred only when it constitutes incitement to discrimination, hostility, or violence (United Nations, 1966, art. 20(2)).
Direct and public incitement to commit genocide is narrower still. It is an international crime requiring a public communication that directly encourages genocidal acts and is accompanied by the necessary genocidal intent. Discriminatory abuse, historical denial, dehumanizing descriptions, or support for persecution may be relevant to a wider factual assessment, but none automatically satisfies that criminal standard.
Broad domestic definitions can weaken the protection of lawful expression. Offenses based on insult, disrespect, offense, or social disharmony may give authorities extensive discretion without identifying the required intent or threatened harm. Political criticism, religious debate, historical research, journalism, satire, and artistic expression may then be punished because they provoke hostility or discomfort rather than because they incite unlawful conduct.
International human rights law does not protect only agreeable or socially accepted ideas. Prejudiced expression can contribute to discrimination and exclusion, but that harm does not eliminate the need to apply the correct legal threshold. The state must identify the rule being invoked and establish each element required by that rule.
1.2 Prohibited, Restrictable, and Protected Speech
Hateful expression can be organized into three legal categories. The first includes speech that states are required to prohibit under specific treaty provisions. Examples include advocacy of national, racial, or religious hatred that constitutes incitement under Article 20(2) of the ICCPR, conduct addressed by Article 4 of the International Convention on the Elimination of All Forms of Racial Discrimination, and direct and public incitement to commit genocide.
The second category covers expression that states may restrict, even though no treaty provision expressly requires its prohibition. Such interference must satisfy the limitations regime governing freedom of expression. Under Article 19(3) of the ICCPR, the restriction must be provided by law, pursue an authorized legitimate aim, and be necessary and proportionate (United Nations, 1966, art. 19(3); Human Rights Committee, 2011).
The third category consists of expression that cannot be prohibited merely because it is hateful, prejudiced, or offensive. A restriction might still be justified on another independently established ground, such as a credible threat, targeted harassment, defamation, or an unlawful invasion of privacy. The state must prove that separate legal basis rather than treating the hate-speech label as sufficient.
Non-coercive responses remain available when state prohibition would be unjustified. Public condemnation, education, counterspeech, victim support, media literacy, and measures addressing discriminatory conduct may respond to harmful expression without imposing criminal liability or compulsory removal.
Private platform rules operate differently. Social-media companies commonly prohibit broader categories of identity-based abuse, degrading language, and harassment than international law requires states to outlaw. Content may violate a company’s community standards while remaining protected against state suppression. Conversely, compliance with platform policy does not establish that a state has discharged its duty to respond to serious threats or unlawful incitement (ARTICLE 19, 2023).
2. Treaty Duties to Prohibit Incitement
International treaties impose several affirmative duties concerning harmful expression, but they do not create one uniform prohibition. The relevant provisions differ in their protected groups, mental elements, anticipated harms, and required forms of domestic implementation.
Article 20(2) of the ICCPR requires prohibition by law but does not prescribe criminal punishment in every case. CERD Article 4 contains more detailed obligations concerning racist expression and organized racist activity. The Genocide Convention establishes individual criminal responsibility for direct and public incitement to commit genocide. Each provision must be interpreted within its own treaty framework.
2.1 Advocacy of Hatred Under ICCPR Article 20(2)
Article 20(2) requires states to prohibit by law any advocacy of national, racial, or religious hatred that constitutes incitement to discrimination, hostility, or violence. Its elements are cumulative. The expression must involve advocacy, concern one of the listed forms of hatred, and amount to incitement toward at least one of the prohibited consequences (United Nations, 1966, art. 20(2)).
Advocacy implies active support or promotion rather than the mere expression of an unfavorable opinion. Incitement requires a connection between the communication and the discriminatory, hostile, or violent response being encouraged. Offensive language about nationality, race, or religion does not satisfy Article 20(2) unless the incitement threshold is also met.
The provision does not specify a single test for likelihood or imminence. The Rabat Plan of Action proposes a contextual assessment based on the surrounding circumstances, the speaker’s position, intent, content and form, the extent of dissemination, and the likelihood and imminence of harm (OHCHR, 2013). Although non-binding, the Rabat framework helps distinguish serious incitement from expression that remains within Article 19.
Intent requires careful treatment. Article 20(2) does not expressly state a mental element, but international guidance generally treats intent to advocate hatred and incite a prohibited response as necessary for serious sanctions. Negligent, careless, or ambiguous expression should not be equated automatically with deliberate incitement.
The provision lists national, racial, and religious hatred. It does not expressly mention sex, gender, sexual orientation, disability, political opinion, or other status. Abuse directed at people on those grounds may engage equality, privacy, security, and non-discrimination guarantees, as well as domestic law. It should not be attributed directly to Article 20(2) without explaining the limits of its text.
“Prohibited by law” does not require states to impose criminal penalties in every instance. In Rabbae v. the Netherlands, the Human Rights Committee confirmed that Article 20(2) does not prescribe a particular form of prohibition (Human Rights Committee, 2016, para. 10.4). Civil, administrative, and criminal measures may all form part of implementation, provided that the response is effective and consistent with the Covenant.
Articles 19 and 20 are complementary. Measures adopted to implement Article 20 must still comply with legality, necessity, and proportionality. A state cannot rely on the mandatory character of Article 20(2) to enact vague offenses or impose sanctions extending beyond the expression that the provision requires it to prohibit (Human Rights Committee, 2011, paras. 50–52).
2.2 Racial Hate Speech Under CERD Article 4
Article 4 of CERD requires states parties to condemn propaganda and organizations based on ideas or theories of racial superiority. Its speech-related obligations include making punishable the dissemination of ideas based on racial superiority or hatred, incitement to racial discrimination, acts of violence against racial or ethnic groups, incitement to such violence, and assistance to racist activities, including financing (United Nations, 1965, art. 4(a)).
Article 4(b) separately requires action against organizations and organized propaganda activities that promote and incite racial discrimination. The provision reaches conduct beyond individual statements and addresses the institutional structures through which racist ideas and activities may be organized.
CERD’s protected grounds derive from its definition of racial discrimination: race, color, descent, and national or ethnic origin. Its obligations are more prescriptive than those in Article 20(2) of the ICCPR in several respects, but their subject matter is narrower. Religious hatred, for example, falls within CERD only when it is connected to racial, ethnic, descent-based, or national-origin discrimination.
Implementation is subject to the provision’s due-regard clause. States must respect the principles of the Universal Declaration of Human Rights and the rights expressly protected by Article 5 of CERD, including freedom of opinion and expression. The clause does not make Article 4 optional, but it prevents criminalization duties from being interpreted without regard to competing human rights.
The Committee on the Elimination of Racial Discrimination has sought to reconcile these obligations in General Recommendation No. 35. Criminalization should be reserved for serious cases, while less severe expression may justify civil, administrative, educational, or other responses. Intent, the social and political context, the speaker’s position, the reach of the communication, and the likelihood of discriminatory action are relevant to determining whether expression warrants criminal punishment (Committee on the Elimination of Racial Discrimination, 2013, paras. 12 and 15–16).
Academic research, journalism, political debate, and discussion of racial or ethnic questions do not become unlawful merely because they reproduce offensive language or examine racist ideas. Context and purpose remain essential. A prohibition intended to combat racial discrimination cannot be applied as a general restriction on controversial historical, political, or social inquiry.
2.3 Direct and Public Incitement to Genocide
Article III(c) of the Genocide Convention makes direct and public incitement to commit genocide punishable. The offense is complete when the prohibited incitement occurs. Proof that genocide was subsequently carried out is not required (United Nations, 1948, art. III(c)).
Directness is interpreted in context. The speaker need not use the legal term “genocide” or describe the intended crime with formal precision. The communication must nevertheless be understood by its audience as encouraging acts intended to destroy a protected group. Language, culture, political conditions, prior communications, and coded references may influence that interpretation.
The incitement must also be public. Speeches to crowds, radio broadcasts, publications, and communications directed toward a broad audience may satisfy that requirement. Online statements can be public when they are accessible or distributed beyond a genuinely private exchange. The platform used is relevant, but audience and accessibility are more significant than the technical form of the message.
Genocidal intent is indispensable. The accused must intend to encourage conduct directed at the destruction, in whole or in part, of a national, ethnic, racial, or religious group as such. Where the alleged objective concerns only part of a group, international criminal jurisprudence requires that part to be substantial. That substantiality requirement comes from judicial interpretation; the Convention itself uses the words “in whole or in part.”
The International Criminal Tribunal for Rwanda developed the offense in Prosecutor v. Akayesu and the Media Case, Prosecutor v. Nahimana, Barayagwiza and Ngeze. The judgments treated directness as dependent on the message’s meaning within its linguistic and cultural setting and confirmed that incitement may be punished even where the intended genocide does not occur (ICTR, 1998; ICTR Appeals Chamber, 2007).
Dehumanizing rhetoric, denial of earlier atrocities, and inflammatory propaganda may provide evidence of intent, context, or an organized campaign against a group. They are not automatically direct and public incitement to genocide. Expanding Article III(c) beyond its elements would weaken the precision required for an international crime and obscure other legal rules that may apply.
3. Restrictions Under Freedom of Expression Law
Many statements described as hate speech fall outside provisions requiring prohibition. States may restrict some of that expression, but they must justify the interference under the applicable law of freedom of expression. Under the ICCPR, Article 19(3) provides the governing framework.
The test applies to criminal convictions, civil judgments, administrative penalties, compulsory removal orders, blocking measures, and comparable exercises of public power. A state must establish a lawful basis, an authorized aim, and a measure that is both necessary and proportionate.
3.1 Legality and Legitimate Aim
A restriction must be prescribed by law. This requirement is not satisfied merely because a domestic statute exists. The rule must be accessible, sufficiently precise, and capable of guiding individual conduct. It must also constrain official discretion and provide protection against arbitrary or discriminatory enforcement (Human Rights Committee, 2011, paras. 24–25).
Terms such as insult, offense, disrespect, hostility, and social disharmony can create problems when left undefined. They may not identify the required mental state, the threatened harm, or the connection between the expression and that harm. The same provision may then be used against direct calls for violence and legitimate criticism of government policy.
Article 19(3) recognizes two categories of legitimate aims: respect for the rights or reputations of others, and protection of national security, public order, public health, or morals. The list is exhaustive. Political convenience, institutional prestige, national unity, and protection of public officials from embarrassment are not independent grounds for restricting speech.
Equality and human dignity may inform the protection of the rights of others, especially where expression contributes to intimidation, exclusion, or discriminatory conduct. They do not remove the need to identify the affected right and explain how the restriction protects it.
International human rights law also distinguishes attacks on individuals or groups from criticism of governments, religions, institutions, and ideas. People are entitled to protection against unlawful discrimination and serious abuse. Belief systems, public authorities, flags, and historical narratives do not possess an equivalent right to be insulated from criticism, satire, or rejection.
3.2 Necessity, Proportionality, and Political Abuse
A legitimate aim does not establish that intervention is necessary. The state must show a specific need for the restriction and a sufficiently close relationship between the expression, the anticipated harm, and the measure imposed. General appeals to social tension, offense, or possible unrest do not meet that burden.
Proportionality requires attention to the severity of the response. Criminal prosecution, civil liability, administrative fines, removal orders, blocking measures, and account restrictions impose different burdens. The speaker’s intent, the seriousness and likelihood of harm, the audience, the reach of the communication, and the availability of less restrictive alternatives all affect the assessment.
Criminal punishment demands the closest scrutiny. The Committee on the Elimination of Racial Discrimination has advised that criminalization should be reserved for serious cases (Committee on the Elimination of Racial Discrimination, 2013, para. 12). Because imprisonment imposes an especially severe interference with expression, its use requires strict justification based on the gravity of the incitement, the speaker’s intent, the likelihood of harm, and the inadequacy of less restrictive measures.
Civil and administrative measures may be less severe, but they are not automatically proportionate. Excessive damages, recurring fines, accelerated removal deadlines, or broad blocking orders can deter lawful expression as effectively as a criminal conviction. Reasoned decisions, access to review, and remedies against erroneous restrictions are necessary safeguards.
Broad speech offenses are particularly vulnerable to political misuse. Governments may invoke hate speech, blasphemy, insult, cybercrime, or public-order laws against journalists, opposition figures, minority advocates, religious critics, and human rights defenders. The Human Rights Committee has emphasized that public institutions and officials are legitimately subject to criticism and that laws protecting them from disrespect require careful scrutiny (Human Rights Committee, 2011, paras. 23 and 38).
Effective protection against discrimination does not require granting public authorities an unrestricted power to suppress hostile or unpopular ideas. The legal response must address a demonstrated harm, remain confined to an authorized purpose, and preserve the safeguards that prevent protection from becoming censorship.
4. Applying the Rabat Threshold Online
The Rabat Plan of Action provides an influential framework for deciding when advocacy of hatred reaches the incitement threshold under Article 20(2) of the ICCPR. It directs attention to six factors: the social and political context, the speaker’s position, intent, content and form, the extent of dissemination, and the likelihood and imminence of harm (OHCHR, 2013, para. 29).
Rabat is not a treaty, judicial decision, or independent source of legal obligation. It emerged from expert consultations convened by the Office of the UN High Commissioner for Human Rights and should be treated as non-binding interpretive guidance. Its six factors do not operate as a points system. Their significance depends on the circumstances, and no fixed number of satisfied factors proves unlawful incitement.
4.1 Context, Speaker, Intent, and Content
Context may include recent violence, entrenched discrimination, armed conflict, electoral unrest, or organized hostility toward an identifiable group. Such conditions do not make expression unlawful on their own, but they affect how a message is understood and how likely it is to produce a prohibited response.
The speaker’s authority may also change the risk. Political leaders, religious figures, military commanders, prominent broadcasters, and influential online personalities can mobilize audiences more readily than ordinary users. Status is relevant because of that capacity, not because prominent speakers receive less protection as a matter of principle.
Intent must be assessed from the communication and its surrounding circumstances. Timing, repeated statements, audience selection, prior conduct, and participation in a coordinated campaign may support an inference of deliberate incitement. Ambiguity should not be resolved solely against the speaker, particularly where severe penalties are contemplated.
Online content often depends on cultural and linguistic knowledge. Memes, coded references, irony, altered images, reclaimed slurs, and community-specific expressions may communicate meanings that are not evident from the words alone. Reposting or quoting hateful language may endorse it, but it may also document abuse, report on an event, or challenge the original message. Context is indispensable when distinguishing incitement from journalism, research, satire, or counterspeech.
4.2 Reach, Likelihood, and Imminence of Harm
The extent of dissemination concerns the audience actually reached or deliberately targeted, not merely the technical availability of a post. Relevant considerations include the platform used, duration of availability, frequency of repetition, audience composition, and the ease with which the content can be redistributed.
Follower counts and engagement figures may indicate reach, but neither establishes incitement. A message directed to a small group prepared to act can present a greater risk than material viewed by a large and largely passive audience. Paid targeting, closed networks, and delivery to users selected by identity, location, or political affiliation may also affect the assessment.
Article 20(2) requires a relationship between advocacy of hatred and incitement to discrimination, hostility, or violence. The likelihood analysis should identify a plausible and sufficiently close connection between the expression and the anticipated harm. Rabat includes imminence within that inquiry, but the precise temporal requirement remains dependent on the type of harm and the applicable legal framework. Remote speculation cannot justify serious criminal or administrative sanctions.
4.3 Speed, Persistence, and Coordinated Amplification
Digital distribution can magnify harm through rapid replication, persistent availability, recommendation systems, and circulation across multiple services. Content may continue to spread through screenshots or copied posts after the original has been removed. These features affect reach and likelihood, but they do not create a presumption that online expression is unlawful.
Algorithmic amplification raises questions that are separate from the legality of the original statement. A recommendation system may expose inflammatory content to a much larger audience than the speaker could reach independently. The resulting harm may reflect both the user’s conduct and the platform’s design choices.
Coordinated campaigns require similar differentiation. Synchronized posting, shared instructions, repeated targeting, or deliberate attempts to evade moderation may show that individual messages form part of organized conduct. Responsibility for creating content, arranging its distribution, recommending it, or profiting from its circulation must be assessed separately. International human rights law does not impose one universal liability rule covering every actor in that chain.
5. Regional Standards and Selected Cases
Regional human rights systems use different treaty texts and judicial methods when addressing hateful expression. European, inter-American, and African standards may influence one another, but each remains tied to its own legal framework. Regional doctrine should not be presented as a universal rule of international law.
5.1 Europe: Articles 10 and 17 of the ECHR
Article 10 of the European Convention on Human Rights protects freedom of expression while permitting restrictions that are prescribed by law, pursue a legitimate aim, and are necessary in a democratic society. The European Court of Human Rights examines proportionality and allows states a margin of appreciation whose breadth varies with the speech, the context, and the quality of domestic review.
Article 17 follows a different route. It prevents Convention rights from being invoked to destroy the rights and values protected by the Convention. The Court has sometimes applied it to exclude claims involving explicitly racist, antisemitic, or Holocaust-denial expression before conducting a full Article 10 proportionality analysis.
In Garaudy v. France, the Court treated Holocaust denial and revisionism as incompatible with the Convention’s fundamental values and declared the application inadmissible under Article 17 (ECtHR, 2003a). That approach has not been applied to every controversial historical claim or intolerant statement. Article 17 remains a selective and case-specific doctrine.
The Court used Article 10 in Gündüz v. Turkey, which concerned statements made by the leader of an Islamic group during a live television debate. Defending sharia without calling for violence was not, in the circumstances, enough to classify the expression as hate speech, and the conviction violated Article 10 (ECtHR, 2003b).
Reliance on Article 17 can prevent the Convention from being used to defend openly destructive projects. It may also remove the opportunity for close analysis of intent, context, harm, and proportionality. The Human Rights Committee has taken a different approach to laws penalizing opinions about historical facts, requiring restrictions to satisfy Articles 19 and 20 of the ICCPR rather than recognizing a general category excluded from protection (Human Rights Committee, 2011, para. 49).
5.2 The Americas and Africa
Article 13 of the American Convention on Human Rights strongly protects expression and generally prohibits prior censorship. In The Last Temptation of Christ v. Chile, the Inter-American Court held that a ban on exhibiting a film violated Article 13, confirming the regional system’s restrictive approach to prior restraints (IACtHR, 2001, paras. 71–73 and 76–80).
Article 13(5) separately requires propaganda for war and advocacy of national, racial, or religious hatred that incites lawless violence or similar unlawful action to be treated as punishable offenses. Its wording is narrower than Article 20(2) of the ICCPR because it focuses on lawless violence or comparable action rather than listing discrimination and hostility as distinct consequences (American Convention on Human Rights, 1969, art. 13(5)).
The African Charter protects expression under Article 9 alongside equality, dignity, and personal integrity. Domestic law cannot use the phrase “within the law” to remove the substance of the right. Restrictions remain subject to legitimate purpose, necessity, and proportionality.
In Lohé Issa Konaté v. Burkina Faso, the African Court found that imprisonment and other sanctions imposed for defamation violated the Charter’s freedom-of-expression guarantees (African Court, 2014). The case did not concern hate speech, but it confirms that serious penalties for expression require close proportionality review. Equality and protection against discriminatory harm must be pursued without granting states an unlimited power to criminalize hostile or unpopular speech.
6. States, Platforms, and Automated Moderation
Human rights treaties principally bind states. Social-media platforms remain private actors, although their rules, ranking systems, and enforcement decisions shape access to information and participation in public debate. Binding state obligations must be distinguished from corporate responsibilities, domestic regulation, contractual commitments, and voluntary policies.
6.1 State Protection and Indirect Censorship
States may have positive obligations to protect individuals against credible threats, serious discriminatory harassment, and incitement by private actors. The scope of those duties depends on the right engaged, the severity and foreseeability of the harm, the authorities’ knowledge, and the measures reasonably available.
Protection does not require removal of every hateful statement. Investigation, evidence preservation, protection of threatened persons, enforcement against discriminatory conduct, and action against serious incitement may be more appropriate than compulsory deletion.
Regulation can also encourage excessive removal. Vague duties to eliminate “harmful” content, strict intermediary liability, and short compliance deadlines may lead platforms to suppress lawful material rather than risk penalties. This shifts censorship decisions to private companies while reducing transparency and judicial oversight.
An independently adopted platform decision is not ordinarily attributable to the state merely because it affects expression. A governmental order or coercive demand may engage the state’s human rights responsibility, depending on the degree of pressure, the connection between the authorities and the restriction, and the legal framework governing the interference.
6.2 Platform Rules and Human Rights Responsibility
Community standards commonly prohibit a broader range of identity-based insults, degrading stereotypes, and harassment than states are required or permitted to criminalize. These policies arise from contracts, product design, commercial decisions, and applicable domestic law. They are not direct applications of the ICCPR.
The UN Guiding Principles on Business and Human Rights state that companies should respect internationally recognized rights, avoid causing or contributing to adverse impacts, address harms directly linked to their services, and provide or cooperate in remediation where appropriate (United Nations, 2011, principles 11–24). The Guiding Principles are non-binding and do not place platforms under the same treaty obligations as states.
Human rights due diligence should examine moderation, recommendation, advertising, and account-enforcement systems. Meaningful assessment requires consultation with affected communities, attention to regional and linguistic conditions, action to reduce identified risks, and evaluation of whether those measures work.
Platform decisions can harm users in opposite ways. Failure to act may expose targeted individuals to intimidation or violence. Over-enforcement may suppress reporting, counterspeech, minority advocacy, or evidence of abuse. Clear rules, reasoned decisions, accessible appeals, and meaningful human review are necessary on both sides of that risk.
6.3 Automation, Language, and Discriminatory Error
Automated tools are used to identify, rank, flag, restrict, and remove content at a scale that manual review cannot match. Their accuracy is limited where classification depends on cultural knowledge, speaker relationships, intent, or current political events.
Dialect, irony, satire, coded speech, reclaimed slurs, and counterspeech produce recurring errors. A system may remove a minority user’s discussion of discrimination while failing to recognize an indirect attack that avoids known prohibited terms. Low-resource languages face greater difficulty when training data, specialist reviewers, and appeal mechanisms are limited.
The European Union Agency for Fundamental Rights found that automated moderation can miss hateful content and produce discriminatory outcomes. Its 2023 study also emphasized that human review is not a complete solution because reviewers may disagree, lack context, or reproduce bias (FRA, 2023).
Moderation errors include both wrongful retention and wrongful removal. Recommendation systems add a separate concern because they determine which material receives prominence. A platform may classify a post correctly under its rules while contributing to harm by repeatedly promoting it to a receptive audience. Content moderation and content amplification require related but distinct assessments.
7. Liability, Remedies, and Procedural Safeguards
Responsibility for online hate speech depends on the conduct and authority of each actor. A speaker creates the message, an account holder may control the forum in which it appears, a platform determines its distribution, and public authorities may order or pressure its removal. These roles cannot be assessed through a single liability rule. The remedy must also correspond to the harm: deletion may stop further circulation, but it may not protect the target, preserve evidence, identify coordinated activity, or provide compensation.
7.1 Third-Party Comments and Intermediary Liability
The European Court of Human Rights has considered when website operators and social-media account holders may bear responsibility for comments posted by other users. Its approach examines the nature of the comments, the operator’s control, the commercial or political setting, the measures available against the original authors, and the speed and effectiveness of any response.
In Delfi AS v. Estonia, the Grand Chamber found no violation of Article 10 after a commercial news portal was held liable for clearly unlawful comments posted beneath one of its articles. The Court emphasized the extreme character of the comments, the portal’s economic interest in user engagement, its control over the comment section, the anonymity of the authors, and the failure of its filtering and notice systems to remove the material promptly. The judgment was expressly confined to a professionally managed news portal and did not establish equivalent duties for every discussion forum or social-media service (ECtHR, 2015).
A different result followed in Magyar Tartalomszolgáltatók Egyesülete and Index.hu Zrt v. Hungary. The comments were offensive and vulgar, but they were not comparable to the threats and clearly unlawful hate speech examined in Delfi. The domestic courts had imposed objective liability without adequately assessing the context, the applicants’ moderation arrangements, the conduct of the injured company, or the effect of liability on online debate. The European Court found a violation of Article 10 and recognized that notice-and-takedown procedures may provide an appropriate response in many circumstances (ECtHR, 2016).
Sanchez v. France concerned a politician who maintained a publicly accessible Facebook wall during an election campaign. The Grand Chamber accepted his criminal liability for failing to delete unlawful comments posted by third parties. It considered the electoral context, the applicant’s choice to open the page to public comments, his political position, the content of the continuing discussion, and his failure to act promptly. The Court did not require automatic filtering or constant prior monitoring; it assessed what vigilance could reasonably be expected from the account holder in the circumstances (ECtHR, 2023).
These judgments do not create a general obligation to monitor every comment before publication. Liability is more readily justified where the content is manifestly unlawful, the intermediary exercises meaningful control, the discussion forms part of a professional or political activity, and the original authors cannot be pursued effectively. Resources, purpose, knowledge, foreseeable risk, and the time taken to respond remain relevant.
The three cases belong to the European Convention system. They may guide comparative reasoning, but they do not constitute a universal rule of intermediary liability under international human rights law.
7.2 Effective and Proportionate Responses
A proportionate remedial framework may include removal, reduced distribution, warnings, limits on comments, temporary account restrictions, preservation of evidence, civil remedies, criminal prosecution, counterspeech, and direct support for targeted users. The appropriate measure depends on the legal classification of the expression, the speaker’s intent, its reach, the likelihood of harm, and the consequences of continued circulation.
Removal is most readily justified where the content is clearly unlawful and its continued availability creates a serious risk. Evidence should be preserved where the material may support an investigation, civil claim, criminal prosecution, or documentation of coordinated abuse. Immediate deletion without preservation can make later accountability more difficult.
Some harms arise mainly from amplification. Limiting recommendation, disabling monetization, attaching a warning, or restricting comments may reduce exposure without erasing the original material. These measures still affect expression and should not operate indefinitely or without explanation. Hidden downranking is particularly difficult to contest because users may not know that their reach has been restricted.
Account-level measures should ordinarily rest on an individualized assessment of the user’s conduct, repetition, intent, and risk of further harm. Coordinated abuse, repeated violations, or deliberate evasion of enforcement may justify suspension. Permanent exclusion requires stronger reasons, especially where the account is used for journalism, political participation, professional activity, or access to a community.
Civil remedies may provide compensation, correction, or an injunction without resort to criminal law. Their proportionality depends on the amount awarded, the breadth of the order, procedural fairness, and the possible chilling effect on lawful expression. Criminal prosecution should remain confined to conduct satisfying a clear offense and serious enough to justify the coercive power of the state.
The interests of targeted users extend beyond the status of the post. They may require protection from credible threats, assistance in documenting abuse, psychological support, legal advice, or access to law-enforcement and equality institutions. Counterspeech, public condemnation, education, and media literacy can also address prejudiced expression that does not justify compulsory removal or punishment.
7.3 Notice, Reasons, Appeals, and Independent Review
Procedural safeguards reduce arbitrary and inconsistent enforcement. Rules should be accessible, sufficiently precise, and available in the languages used on the service. Users need to know whether a restriction arises from domestic law, a governmental order, or the platform’s own community standards.
A person whose content is removed, demonetized, restricted, or made less visible should ordinarily receive prompt notice. The explanation should identify the content affected, the rule applied, the measure imposed, and the principal reason for the decision. A generic reference to “community standards” rarely provides enough information for a meaningful challenge.
Human review is especially valuable where a decision depends on irony, cultural meaning, quotation, political context, or counterspeech. Review should involve more than automatic confirmation of the first result. The reviewer must have access to the relevant context, the user’s explanation, and the authority to reverse the decision.
Appeal procedures should be accessible and capable of producing a timely remedy. Restoration after an election, protest, breaking-news event, or urgent public debate may come too late to repair the interference. Appeals should cover account sanctions and distribution restrictions as well as complete removal.
Where public authorities compel or strongly pressure a platform to restrict expression, access to independent review becomes especially important and may be required under the applicable human rights framework. A court, regulator, or other reviewing body must possess sufficient independence and authority to assess the legal basis, evidence, and proportionality of the restriction.
Transparency reporting should distinguish enforcement under private rules from action taken pursuant to legal orders or government requests. Useful reporting also addresses the types of measures imposed, languages and jurisdictions affected, use of automation, appeals, reversal rates, and patterns of wrongful removal or retention.
Remedies must operate in both directions. Users need redress when lawful expression is removed or repeatedly restricted in error. Those exposed to serious threats or incitement need procedures capable of securing prompt action, preserving evidence, and addressing continuing danger. A system concerned only with censorship neglects the rights of targeted users; one concerned only with harmful content invites indiscriminate removal.
Also read
Conclusion
International human rights law does not establish a general prohibition of online hate speech. It requires states to prohibit defined forms of advocacy and incitement, while other restrictions must satisfy legality, legitimate aim, necessity, and proportionality. Expression falling below those thresholds cannot be suppressed merely because it is prejudiced, offensive, or disturbing.
Digital communication changes the evidence relevant to that assessment. Speed, persistence, targeting, coordinated activity, and algorithmic recommendation may increase reach and the likelihood of harm. They do not dispense with the need to establish intent, context, and the elements of the applicable legal rule.
Effective regulation depends on accurate classification and accountable decision-making. State duties, regional judgments, domestic laws, platform policies, and corporate human rights standards remain legally distinct even when they address the same content. Precise rules, proportionate responses, reasoned decisions, and effective review provide stronger protection than either indiscriminate censorship or an abstract appeal to freedom of expression.
References
African Charter on Human and Peoples’ Rights (1981) adopted 27 June 1981, entered into force 21 October 1986, OAU Doc. CAB/LEG/67/3 rev. 5, 1520 UNTS 217.
African Court on Human and Peoples’ Rights (2014) Lohé Issa Konaté v. Burkina Faso, Application No. 004/2013, judgment on the merits, 5 December 2014.
American Convention on Human Rights (1969) adopted 22 November 1969, entered into force 18 July 1978, OAS Treaty Series No. 36, 1144 UNTS 123.
ARTICLE 19 (2023) Content Moderation and Freedom of Expression Handbook. London: ARTICLE 19. Available at: https://www.article19.org/wp-content/uploads/2023/08/SM4P-Content-moderation-handbook-9-Aug-final.pdf (Accessed: 01 July 2026).
Clooney, A. and Gardoll, A. (2024) ‘Hate speech’, in Clooney, A. and Neuberger, D. (eds.) Freedom of Speech in International Law. Oxford: Oxford University Press, pp. 153–219. doi: 10.1093/law/9780198899372.003.0003.
Committee on the Elimination of Racial Discrimination (2013) General Recommendation No. 35: Combating Racist Hate Speech, CERD/C/GC/35, 26 September 2013.
Convention for the Protection of Human Rights and Fundamental Freedoms (European Convention on Human Rights) (1950) adopted 4 November 1950, entered into force 3 September 1953, ETS No. 5, 213 UNTS 221.
Convention on the Prevention and Punishment of the Crime of Genocide (1948) adopted 9 December 1948, entered into force 12 January 1951, 78 UNTS 277.
European Court of Human Rights (2003a) Garaudy v. France, Application No. 65831/01, decision on admissibility, 24 June 2003.
European Court of Human Rights (2003b) Gündüz v. Turkey, Application No. 35071/97, judgment, 4 December 2003.
European Court of Human Rights (2015) Delfi AS v. Estonia [GC], Application No. 64569/09, judgment, 16 June 2015, ECHR 2015.
European Court of Human Rights (2016) Magyar Tartalomszolgáltatók Egyesülete and Index.hu Zrt v. Hungary, Application No. 22947/13, judgment, 2 February 2016.
European Court of Human Rights (2023) Sanchez v. France [GC], Application No. 45581/15, judgment, 15 May 2023.
European Union Agency for Fundamental Rights (2023) Online Content Moderation: Current Challenges in Detecting Hate Speech. Luxembourg: Publications Office of the European Union. doi: 10.2811/923316. Available at: https://fra.europa.eu/sites/default/files/fra_uploads/fra-2023-online-content-moderation_en.pdf (Accessed: 05 July 2026).
Griffin, R. (2023) ‘Rethinking rights in social media governance: human rights, ideology and inequality’, European Law Open, 2, pp. 30–56. doi: 10.1017/elo.2023.7.
Human Rights Committee (2011) General Comment No. 34: Article 19—Freedoms of Opinion and Expression, CCPR/C/GC/34, 12 September 2011.
Human Rights Committee (2016) Mohamed Rabbae, A.B.S. and N.A. v. the Netherlands, Communication No. 2124/2011, Views adopted 14 July 2016, CCPR/C/117/D/2124/2011.
Inter-American Court of Human Rights (2001) The Last Temptation of Christ (Olmedo Bustos et al.) v. Chile, merits, reparations and costs, judgment, 5 February 2001, Series C No. 73.
International Convention on the Elimination of All Forms of Racial Discrimination (1965) adopted 21 December 1965, entered into force 4 January 1969, 660 UNTS 195.
International Covenant on Civil and Political Rights (1966) adopted 16 December 1966, entered into force 23 March 1976, 999 UNTS 171.
International Criminal Tribunal for Rwanda (1998) Prosecutor v. Jean-Paul Akayesu, Case No. ICTR-96-4-T, Trial Chamber judgment, 2 September 1998.
International Criminal Tribunal for Rwanda (2007) Prosecutor v. Ferdinand Nahimana, Jean-Bosco Barayagwiza and Hassan Ngeze, Case No. ICTR-99-52-A, Appeals Chamber judgment, 28 November 2007.
McGonagle, T. (2013) The Council of Europe Against Online Hate Speech: Conundrums and Challenges. Expert Paper MCM(2013)005. Strasbourg: Council of Europe. Available at: https://rm.coe.int/090000168059bfce (Accessed: 31 July 2026).
Office of the United Nations High Commissioner for Human Rights (2013) Report of the United Nations High Commissioner for Human Rights on the Expert Workshops on the Prohibition of Incitement to National, Racial or Religious Hatred, A/HRC/22/17/Add.4, 11 January 2013. Available at: https://digitallibrary.un.org/record/746343 (Accessed: 07 July 2026).
United Nations (2011) Guiding Principles on Business and Human Rights: Implementing the United Nations “Protect, Respect and Remedy” Framework. New York and Geneva: United Nations, HR/PUB/11/04. Available at: https://www.ohchr.org/sites/default/files/documents/publications/guidingprinciplesbusinesshr_en.pdf (Accessed: 23 July 2026).
United Nations Human Rights Council (2016) The Promotion, Protection and Enjoyment of Human Rights on the Internet, A/HRC/RES/32/13, resolution adopted 1 July 2016. Available at: https://digitallibrary.un.org/record/845727 (Accessed: 23 July 2026).
