Why We Should Be Concerned about The “Illegal Hate Speech” Code of Conduct +
The DSA has integrated the Code+ into its framework. Will Europe's attempt to deal with online illegal hate speech undermine free expression?
The European Union’s sweeping online safety law continues to make headlines — some controversial — as we learn more about how it will be implemented and enforced in practice. That law, the Digital Services Act (DSA), has been hailed as a novel “road map” for reigning in the “Wild West” of the internet that has allegedly been overrun with “torrents” of hate speech and illegal content.
But how exactly will these new regulations ensure that platforms deal with illegal hate speech without stifling freedom of expression online? Recent developments raise some significant red flags.
In 2016, the European Commission introduced the Code of Conduct on Countering Illegal Hate Speech Online, a voluntary agreement with major tech companies aimed at ensuring the swift removal of illegal online hate speech. It marked an early attempt at co-regulation in the absence of binding EU-wide legislation.
Over time and with the adoption of the DSA, this initiative evolved into a more formalized and enforceable framework, culminating in the Code of Conduct on Countering Illegal Hate Speech Online + (Code+). In January 2025, this code was integrated into the Digital Services Act (DSA). The signatories to date include Facebook, Instagram, X, YouTube, TikTok, Dailymotion, Jeuxvideo.com, LinkedIn, Microsoft-hosted consumer services, Snapchat, Rakuten Viber, and Twitch.
The first issue of concern is the extent to which the Code+ will, in practice, be voluntary. The DSA explicitly states that the European Commission and the European Board for Digital Services shall “encourage and facilitate the drawing up of voluntary codes of conduct at Union level to contribute to the proper application of this Regulation.” Recital 103 of the DSA notes while codes should be measurable and subject to public oversight “this should not impair the voluntary nature of such codes.”
Nevertheless, Article 45 of the DSA grants the Commission and the European Board for Digital Services oversight powers, allowing them to exert pressure on signatories for Code compliance. Specifically, Part 4 of this article stipulates that “in the case of systematic failure to comply with the codes of conduct, the Commission and the Board may invite the signatories . . . to take the necessary action.”
What exactly constitutes a “necessary action” is never defined, but the infrastructure of the Code+ within the ambit of the DSA could effectively transform ‘voluntary commitments’ into more rigid obligations under regulatory scrutiny. This framework could end up creating a de facto enforcement mechanism, blurring the line between voluntary adherence and regulatory compulsion. Moreover, the DSA requires platforms to mitigate “systemic risks” while the Code+ refers to Article 35 (1) (h) of the DSA, which renders adherence to voluntary codes as an appropriate risk mitigation measure in the framework of Very Large Online Platforms and Very Large Online Search Engines.
In light of the above nuances, and while the Code+ is framed as a voluntary instrument creating “voluntary commitments,” it could be argued that, in practice, its implementation will not pan out to be completely voluntary. While time will tell, particularly in terms of the rigor and findings of the monitoring reports, civil society organizations and other stakeholders must remain vigilant to ensure that the voluntary nature of such codes is retained.
Additionally, several issues that plagued the 2016 Code persist, including the absence of a clear definition of hate speech, the risk of excessive content removal, and the reliance on trusted flaggers. These challenges could become even more problematic with the integration of the Code+ into the DSA framework.
A Lacking Definition of Hate Speech
A key issue with the DSA is its lack of a clear definition of hate speech. The DSA obligates digital platforms to assess and remove illegal content, including hate speech and unlawful discriminatory content, but does not specify what constitutes “illegal content.” Instead, it leaves platforms with considerable discretion, requiring them to rely on national laws aligned with EU regulations or interpret EU law directly.
The Code+ follows the precedent set by the 2016 Code of Conduct, using the Framework Decision on Racism and Xenophobia and national laws, transposing it as a definitional benchmark. However, this reliance presents several challenges. The Framework Decision is a criminal law instrument that sets a high threshold for prohibition, focusing on incitement to violence or hatred. It also does not explicitly define hate speech and only protects characteristics such as race, color, religion, descent, and national or ethnic origin—offering a narrower scope than most social media policies.
Despite the reference to the Framework Decision and national laws, platforms must remove content that violates their policies as well as national laws more broadly and not only those transposing the Framework Decision. However, broad platform policies that extend, amongst many others, to expressions of contempt and stereotypes are often a far cry from international human rights standards. Also, variations in national legal frameworks create uncertainty for platforms operating across Europe.
Given the complexity and contentious nature of hate speech regulation, both the Code+ and the DSA would have benefited from a precise definition. Nevertheless, even with such a definition, tasking private companies that are not bound by International Human Rights Law to ascertain the limits of the fundamental freedom of expression is a serious problem in the current infrastructure.
More Over-Removal?
A key concern arising from the Code+ and its role in the DSA’s ambit is the over-removal of content. It is essential to first acknowledge that the DSA mandates platforms to remove illegal hate speech “without undue delay.” However, the Code+, like its predecessor, imposes a 24-hour deadline for content removal. In terms of quantitative requirements, the signatories commit to review at least 50% of notices received. They will “apply their best efforts to go beyond this target” and aim at least 67% (two-thirds) of the notices.
In a 2021 report, the UN Special Rapporteur on the Promotion and Protection of the Right to Freedom of Opinion and Expression, Irene Khan, cautioned that increased platform liability may incentivize intermediaries to engage in the excessive removal of content. As she poignantly noted, “By compelling social media platforms to police speech, they create a risk that companies will zealously over-remove material and undermine free speech.”
To deal with the sheer amount of online content but also due to enhanced obligations, platforms rely on automated content moderation. As emphasized in a Council of Europe report, automated mechanisms have a direct impact on freedom of expression, raising concerns regarding the rule of law, particularly in terms of legality, legitimacy, and proportionality. The Council of Europe cautioned that the increased reliance on AI for content moderation could lead to over-blocking, thereby endangering freedom of expression. Additionally, in a 2020 study, Gorwa, Binns, and Katzenbach argue that the growing use of AI risks exacerbating the existing opacity of content moderation, complicating online justice, and “re-obscur[ing] the fundamentally political nature of speech decisions being executed at scale.”
In other words, automated mechanisms inherently lack the capacity to grasp the nuance and context of language and human communication. As my colleague Jacob Mchangama pointed out in a previous article, both human and automated content moderation led to the removal of several satirical and irony-filled posts on Facebook and Instagram due to a failure to understand such context.
With the combined pressure to remove content at such a large scale and the potential legal repercussions of failing to do so, it is safe to assume that platforms will inevitably over-remove content. In a 2024 report, The Future of Free Speech found that a significant proportion of content removed by Facebook and YouTube in France, Germany, and Sweden was, in fact, legally permissible. Depending on the dataset, between 87.5% and 99.7% of deleted comments fell within the bounds of legality.
Monitoring Reporters and Trusted Flaggers
The results of the monitoring exercises raise another issue of concern pertaining to “trusted flaggers” and “monitoring reporters.” Both the DSA and the Code + refer to trusted flaggers, which are entities appointed by national authorities and have expertise in detecting, identifying, and notifying illegal content.
The Code + also provides for a network of “monitoring reporters,” which are “not-for-profit or public entities with expertise on illegal hate speech in at least one EU Member State and approved by the Commission and the Signatories to participate in the Monitoring Exercise.” The monitoring reporters will be conducting annual monitoring exercises to ascertain the extent to which the signatories are complying with the Code+.
Monitoring reporters may also be trusted flaggers. Under the original Code, trusted flaggers were primarily equality and non-discrimination bodies, along with some national institutions. Notably, there was an absence of free speech organizations among the trusted flaggers, as can be seen in the lists of flaggers in each of the seven monitoring exercises.
This composition of groups could lead to a bias toward content removal since it creates an imbalanced oversight mechanism that risks undermining fundamental rights. The absence of a free-expression perspective means that decisions about content removal may disproportionately prioritize offense avoidance over legitimate discourse, satire, or political critique, for example. Moreover, free speech organizations play a critical role in scrutinizing whether removals align with principles of legality, necessity, and proportionality, ensuring that restrictions on expression do not erode democratic debate.
It remains to be seen which organizations (at least one per country) will undertake the monitoring exercises, but a better approach would be to adopt a more holistic composition of organizations, especially compared to the 2016 Code. If both non-discrimination and free speech advocates participate, it will create a more balanced, transparent, and rights-respecting content moderation framework. This recommendation would not solve the inherent issues and controversies of the EU’s approach to content moderation, but it could certainly improve the status quo.
The fact that monitoring reporters can also be trusted flaggers appointed under the DSA creates additional issues. Under Article 22 of the DSA, trusted flaggers are appointed by national authorities, which raises concerns about potential governmental overreach, especially in countries with authoritarian tendencies.
For instance, in countries where the government has been criticized for undermining democratic institutions, their appointment of trusted flaggers could lead to the suppression of dissenting voices under the guise of combating hate speech. Moreover, the opacity surrounding how national regulators appoint trusted flaggers makes it challenging to assess whether these actors are impartial and adequately representative.
Counter-Narratives
The Code+ also refers to enhanced cooperation between stakeholders to raise awareness of illegal hate speech, including how to report it and prevent its spread. From a free speech perspective, the inclusion of counter-narratives in the Code+ is particularly significant.
The 2016 Code acknowledged the role of civil society organizations in countering online hatred through the development of alternative narratives and pledged collaboration between signatories and the European Commission to promote independent counter-narratives. However, in practice, the emphasis remained mainly on monitoring and content removal, leaving the implementation of this commitment unclear. With the Code+, the counter-narrative approach has been reaffirmed, and ideally, this time, greater resources and attention will be dedicated to developing meaningful tools and initiatives in this area, which have proven effective for combatting hate speech (online and offline).
Free Expression Needs Consideration
Nowhere does the Code + refer to freedom of expression. This critical absence reflects the deterioration of this fundamental freedom under the current European approach to the moderation of online content.
The Code+’s addition of a timeframe creates additional pressure on platforms to remove content, while the continuation of the monitoring exercise increases the chances that considerations for free expression will be wholly absent from how the code works in practice.
Let’s hope that free expression advocates and experts will have a seat at the table. If not, we should continue to demand that this fundamental right be weighed equally with the other objectives of the DSA as it comes into full focus.
Natalie Alkiviadou is a Senior Research Fellow at The Future of Free Speech. Her research interests lie in the freedom of expression, the far-right, hate speech, hate crime, and non-discrimination.