Why do social media protect keyboard warriors?

With the emergence of social networks as the main spaces for public expression, the phenomenon of online hate speech has taken on increasing legal and social significance. According to the European Commission's definition, hate speech refers to any communicative behaviour that incites violence, hatred or discrimination against individuals or groups on the basis of race, religion, gender, sexual orientation or other personal or social conditions¹.

The transnational nature of digital platforms and the ease of anonymity online make it very difficult to identify the authors of such content. In Italy, prosecutors have been complaining for years about the ineffectiveness of the tools available to obtain identification data from entities such as Meta, X or TikTok in a timely manner². At the same time, the limitations imposed by privacy legislation and the platforms' own terms of use hinder transparency and cooperation with judicial authorities.

We examine the current regulatory and technological context, highlighting how social media platforms' resistance to identifying users is one of the main obstacles to legal protection against online hate speech.
The regulatory framework against online hate speech

In Italy, the Constitution protects freedom of expression (Art. 21), but this right is not absolute and is limited by respect for human dignity and public order. The repression of hate speech is based on criminal laws such as:

Art. 604-bis of the Criminal Code, which punishes propaganda and incitement to commit crimes on grounds of racial, ethnic and religious discrimination;
Art. 595 of the Criminal Code, which regulates defamation, including through electronic means;
Art. 612-bis of the Criminal Code, which punishes acts of persecution, including cyberstalking.
At the supranational level, Article 19 of the International Covenant on Civil and Political Rights protects freedom of expression, but also provides for restrictions necessary “for respect of the rights or reputations of others”.

Difficulties in the Italian and European regulatory framework

The Digital Services Act (EU Reg. 2022/2065), which came into force in 2024, introduces general transparency and accountability obligations for platform providers, requiring very large ones (VLOPs) to operate effective reporting systems, publish transparency reports and cooperate with authorities⁴. However, direct identification of users remains possible only if the platform decides to store and provide the requested data, which rarely happens.

Anonymity and privacy: the main obstacles to the suppression of hate speech

One of the main causes of impunity for perpetrators of hate speech lies in the possibility of using pseudonyms, fictitious accounts or anonymisation tools (cross-border VPNs, Tor network, proxy servers). These tools, although legitimate, make it difficult to track users.

Cooperation between platforms is uneven and often insufficient. In many cases, the Italian authorities receive no response from US legal offices or have to wait months to obtain partial data. As highlighted by an investigation by Il Sole 24 Ore, “big tech companies hide behind privacy, even when judges request information in clear cases of hate speech”.

The legislation on the protection of personal data (EU Regulation 2016/679 – GDPR) provides for the possibility of processing data to fulfil a legal obligation or in the context of legal proceedings. However, many platforms interpret these provisions restrictively, denying access to data unless there is a specific court order, forcing the authorities to resort to lengthy international letters rogatory⁶.
Difficulties in enforcement: between case law and legislative uncertainties

Italian case law has often had to dismiss criminal proceedings due to the impossibility of identifying the authors of hateful content. In a well-known case in 2022, the Court of Milan dismissed a case involving a series of racist comments posted on a Facebook profile under a false name, despite the opening of a file for incitement to racial hatred under Article 604-bis of the Italian Criminal Code.

At the European level, the European Court of Human Rights has also recognised the principle that online anonymity may be restricted for reasons of public safety or the protection of the rights of third parties. However, this principle often remains unimplemented, especially in the absence of effective cross-border cooperation mechanisms.

The case of the Italian environmental journalist threatened on X and Telegram in 2023 – documented by the International Press Institute – shows how, despite reports and complaints, legal action remained blocked due to a lack of information useful for tracing the perpetrators⁸.

Artificial intelligence as a solution for detecting hate speech

Platforms are progressively adopting artificial intelligence (AI) tools to detect illegal content, including hate speech. However, the accuracy of AI models varies and depends on the linguistic and cultural context. According to a 2025 study conducted by Oyelami and Tosin, multilingual models such as XLM-R perform well in contexts with high linguistic resources, but struggle with minor languages or ambiguous content⁹.

Furthermore, hate speech is often implicit, ironic or contextual, making it difficult to automate detection. TikTok's algorithm, for example, has been accused of amplifying divisive content based on engagement metrics, even when it leads to extremism or misinformation¹⁰.

Some start-ups, such as Pindrop, offer tools to track users' voices and identify patterns of disinformation, but these technologies remain marginal and also raise significant concerns in terms of surveillance and privacy¹¹.

Regulatory proposals and solutions to improve the fight against hate speech

To overcome the impasse between privacy and the suppression of online hate, a number of legislative proposals have emerged:

A single digital identity register for social media, linked to a real document (with obvious constitutional issues);
Data retention obligations extended to platforms, to be harmonised at European level to avoid forum shopping;
Bilateral agreements between states and big tech, based on the model of the CLOUD Act between the US and the EU;
National structures for managing data requests, as provided for in the Digital Services Act.

Another option is to encourage assisted self-regulation through codes of conduct and algorithmic auditing mechanisms with the participation of authorities and civil society. Article 35 of the DSA provides for the possibility for national authorities to appoint independent supervisory bodies for platforms¹².

In Italy, some members of parliament have proposed amendments to the law on defamation and the regulations on provider liability (Article 17 of Legislative Decree 70/2003) in order to make cooperation mandatory for criminally relevant content, but the proposals are still under discussion.
Future prospects in the fight against online hate speech

The phenomenon of online hate speech requires a redefinition of legal protection tools, taking into account the fluid, global and algorithmic nature of digital communication. Failure to identify perpetrators is not only an investigative problem, but also a matter of protecting fundamental rights, as it undermines trust in institutions and the possibility of obtaining justice.

Social media platforms must cease to be invisible arbiters of public debate and assume responsibilities commensurate with their power. Without effective tracking, collaboration and transparency mechanisms, the fight against hate speech remains ineffective and, often, merely symbolic.

The balance between anonymity and accountability, privacy and security, expression and responsibility will be the key issue in digital law over the next decade. Only an integrated legal-technical approach, based on interoperability, verification and cooperation, will make it manageable in practice.


Notes
[1] Art. 35 EU Regulation 2022/2065.
[2] European Commission, Code of conduct on countering illegal hate speech online, 2022.
[3] Il Sole 24 Ore, Hate speech on the web, identification hampered by platforms, 2024.
[4] UN, International Covenant on Civil and Political Rights, 1966.
[5] Regulation (EU) 2022/2065, Digital Services Act.
[6] Il Sole 24 Ore, cit.
[7] Italian Data Protection Authority, Guidelines on data processing in judicial proceedings, 2023.
[8] ECHR, judgment of 10 March 2020, Magyar Helsinki Bizottság v. Hungary.