By Rafa Akter

On June 3, 2026, TIU IRSA hosted a Guest Lecture by Professor Juan Espíndola Mata from the National University of Mexico. Professor Juan  Espíndola  (PhD, Michigan) is an Associate Professor at the Institute for Philosophical Research. His research focuses on philosophical aspects of AI and Transitional Justice. Currently, his research interests are in Content Moderation, Social Media Platforms, and Human Rights. He appeared live at TIU Ikebukuro Campus and delivered his presentation on “Dangerous speech and content moderation in digital platforms.” 

This article will summarize the lecture by Professor Juan Espíndola on “Dangerous speech and Content Moderation in Digital Platforms”. It will include how the content moderations are taken into account, the removal of content and its limitations, weaponization of platforms by different organized criminal organizations; specifically cartels in Mexico, and the harms enabled by the platforms. 

There are two facts about hate speech and dangerous speech, or the related categories: first, it is not “innate,” meaning that it does not arise fundamentally, but is taught. Secondly, people are affected by hate speech cumulatively, gradually, and through the dint of repetition (Susan, 2023). To elaborate, the teachings that are intended to implement hatred towards certain groups are repeated consecutively and attempt to internalize the hate among the community. While the concept of hate speech and its impacts remains broad in terms of spreading the teaching to the people, its media of teaching and to the group who are targeted to be hated, this article will specifically focus on online hate speech and digital harassment, following the content moderation due to the existing social media platforms’ policies on what categorizes hate speech and what action has to be taken. To elaborate, moderation refers to the idea of decision making processes of the platforms to decide if content should stay on the media or is to be removed.

The rapid growth of social media usage and its accessibility by the global population has connected individuals, organizations, and businesses beyond geographic limitations, gradually transforming into a powerful driving force for globalization by connecting populations across cultures, nations, and continents. It has challenged the power dynamic of traditional media channels, often controlled by powerful institutions and dominant groups, and has become a seemingly empowered platform that amplifies the voices across regions, among diverse groups and organizations (Kumar et al., 2024). However, the tension arises when this amplified voice comes from criminal organizations, for instance, drug cartels in Mexico weaponizing social media through three means: (i) recruiting, (ii) sowing terror, and (iii) disseminating propaganda. While this kind of content violates the community standards, the most common innate response from the regulatory bodies is to remove the content, scrub the videos, or ban the account. However, the action of removing the content is understandable as a conduct to prevent hate speech, terror, and threats from those organizations, but it seems strategically naive. Professor Juan Espíndola states that this standard of setting acceptable content by platforms tends to shape “one-size-fits-all” standards, which he calls a “fool’s errand” and that raise significant normative concerns. A normative alternative to this is a pluralistic approach that is sensitive to local content. It combines removal, demotion, and labeling; organizes both automated AI detection and human review; deploys salaried or voluntary moderators and crowdsourcing; and assesses content and network levels (Gorwa et al. 2020; Tobi, 2024). 

Before assessing the cartels, Professor Espíndola discussed the broader issues of corporate power and private authority. Platform companies have direct access to a massive number of the global population; for instance, Facebook alone has 3 billion users, which is one-third of the whole global population. The platform companies make rules for the users, make regulations on what speech is acceptable, and set global standards for each platform for governing users across more than 180 countries, acting as a non-state actor exercising rulemaking, similar to transnational bodies, and have become influential security actors in their own right. These platform institutions are not neutral and are influential actors that exercise quasi-governmental authority, playing a dual character by their private ownership and operation, and are often state-integrated with their critical systematic services (Yiğenoğlu & Kızıldere, 2025).

To conceptualize corporate power and private authority as the functionality of these companies, Professor Juan Espíndola refers to the literature of Srivastava (2023), that platforms create collective dependencies as infrastructural power and exercise discretionary power. Its legibility derives from their schemes of collecting user and non-user data and positioning individuals into groupings, which have not been done in any way before. Srivastava (2023) draws platforms as the racketeers, referring to Charles Tilly’s argument, “create a threat and then charge for its reduction” and “To the extent that the threats against which given government protects its citizens are imaginary or are consequences of its own activities, the government has organized a protection racket” (Tilly, 1985:171). Applying this argument of platforms, Srivastava (2023) indicates platforms are playing as a racket by constructing their protection role on the subject of violence that is produced by their own. For instance, the protection racket of Facebook directly connects to its growth strategies and global scale, legitimizes protection rights by generating a “resolving” mechanism of harm created by the increased polarization, rapid spread of harassment and libel, and radicalization, that social media claims to address through their content moderation processes. This dynamic is highly prevalent in the Global South, especially when the deep interventions of platforms constitute chronic underinvestment in content moderation, limited regulatory capacity, and weak corporate oversight. The stakes are analogously higher where the state is highly integrated with criminal governance, for instance, Mexico.

Weaponization of Platforms by Cartels in Mexico

Professor Juan Espíndola explains weaponization of social media platforms into three different categories: recruiting, sowing terror, and disseminating propaganda, and each contains its own complexities in moderation.         

 Recruiting

The first immediate action to weaponize media is using platforms to recruit new members. The cartels in Mexico maintain a “reserve army of drug labor,” which means individual resistance might not be enough because some other individuals will be effectively replaced to run their system. The cartels exercise their labour supply in similar to the way capitalism runs its workforce, which is the collection of underemployed or unemployed individuals that capitalism maintains to discipline the working class. The cartels constantly require low-level workers’ supply to continue their operations, and these workers are positioned in three roles. The first one is the role of lookouts, where the children and teenagers are kept in charge to watch for police or rival cartel movements in the neighborhoods and report them back, offering low wages yet highly disposable. Secondly, the position of mules, individuals who transport drugs physically across the points of drug supply or between the borders, and are recruited often from vulnerable migrant communities or from highly poor families. They are replaced easily if arrested or killed. Thirdly, the hitman roles, known as the assassins or enforcers, who are paid comparatively higher but possess higher risks. The recruitment process on the platforms is the way that cartels replenish the workers at a scale by offering 200-500 USD approximately per month to the economically marginalized youths, and sometimes the recruitment gets deceptive as they advertise the roles as “security guards” on TikTok, Telegram, WhatsApp, and Facebook, designing the posts resembling the legitimate job listings by “Organization of the Four Letters”. These recruitments become competitive in areas with high unemployment, thereby mobilizing a “reserve army of labour” under similar regulation by capitalist firms. 

To tackle this problem, platforms take content moderation into account, removing content and banning accounts that promote criminal group activities and criminal socialization by enforcing policies related to recruitment. In addition, the NLP-based automated detection mechanism of language models also helps identify coded job offer language by analyzing the role description, salary ranges, and recruitment phrasing similar to cartel operations. However, this raises the risk of over-removal. Mexico has a prevalent informal economy and precarious job markets, and many small businesses exist that are not registered with the authorities formally. As a consequence of extreme moderation, this potentially could cause harm to small businesses, workers, and economic activities.

Sowing Terror

Sowing terror by dissemination of graphic violence, the second form of weaponization, seemingly most instinctively disturbing form. The cartel sows terror for three purposes: (i) building panic as warfare by creating widespread fear. This mass fear leverages them for gaining territorial control and social domination over civilian populations; (ii) intimidating rivals by using the graphic content dedicated to the competing rival groups to indicate the consequences of defiance and deterring challenges; (iii) demanding acknowledgment from the government and law enforcement via sending public signals through conducting atrocities. Removal of the contents, as platforms’ action to eradicate these weaponization acts by cartels, automated AI models identify millions of items in a very short time; however, automated content moderators are sometimes inaccurate on their own, remove legitimate journalism, and have higher false positive rates, which puts legitimate content at risk. On the other hand, salaried human moderators produce more accurate judgments by analyzing the context and catching coded language content that sometimes AI misses; however, it causes a significant traumatizing effect on the human moderators, facing serious harm due to the systematic exposure to traumatizing graphic content. These moderators often come from the Global South for cheap labour, usually receive no psychological counselling, and are susceptible to various mental health disorders, such as severe anxiety, PTSD, and empathy fatigue, causing dehumanizing symptoms in them. This wrongful exploitation of employees coming from the Global South who receive no adequate safeguards raises the moral concern of platform companies, intensifying the “North-South” disputes. Nevertheless, reliance on automated moderation potentially risks the removal of human rights documentation atrocities’ evidence due to its context analysis limitations, causing the “Problem of Demobilization”. To elaborate, the reality of massive violations of human rights is concealed from the broader society even when the accurate contents are removed. This promotes active ignorance and the sentiment of “they need not to know” (Medina et al.). As a consequence of this, domestic and international communities may lack moral incentives to take action against violence and impunity. 

Disseminating Propaganda

The third platform weaponization operates at the cultural level. Cartels actively portray themselves as benevolent forces and noble, and this impression is conveyed to people. They shows to perform community services by food and aid distribution, cultivating “ambient glamorization”. Ambient glamorization refers to dignification or glamorization of traffickers’ image on social media, establishing ‘pro-cartel’ aesthetics across society emerging from social media depictions. To illustrate, narcocorrido, a northern Mexican folk musical genre, celebrates loyalty and the cartel identity. The narcocorridos artists present these charismatic depictions of cartels through their music is streamed globally, particularly on YouTube and Spotify. This glamorizes criminal organizations as courageous, desirable, and honorable and legitimizes Meta’s own community standards, which it otherwise prohibits the content of criminal glorification. To elaborate, Meta has a policy that clearly prohibits the action of glorifying dangerous organizations. It states distinctly: “Legitimizing or defending violent acts by claiming they have a moral, political, or logical justification that makes them appear acceptable or reasonable,” and “Characterizing or celebrating the violence or hate of a designated entity as an achievement or accomplishment” (Meta, 2022). The glorification of boss figure cartels by Narcocorrios raises the normative concern if Meta should cover its policy on this content.

Additionally, algorithmic amplification, the effect of platforms, widens the propaganda. For instance, TikTok expands its audience that spreads the glamorization of narcolife on a large scale. YouTube advocates the videos of narcocorridos and aid distribution by cartels to audiences beyond the traditional radio reach, intensifying exposure and enhancing radicalization. Facebook connects sustained cartel supporter communities over time, a platform that builds community by connecting individuals and groups of the same kind. The further normative objection to removal mechanisms of platforms derives from the concern of eliminating items that are legitimate scholarly interests, such as cartel media and narcocorridos, which are subjects to academic study, and eliminating content reflects the removal of sources to conduct the study, undermining critical thinking because exposure of propaganda to the users enables acknowledgment, analysis, and refutation. Additionally, legitimate and journalistic content may get suppressed, and valuable information may get censored, which emerges as the normative objection to the removals.

                                                          Lecture Takeaway

While platform companies set universal standards across borders, nationalities, and the sovereignty of each nation, it can be hypocritical actor due to the racketeer problem and morally objectionable because of their moderation labor exploitation. A system that potentially suppresses journalistic and academic materials, demobilizes civil society, traumatizes moderators with no accountability and incentives, when applies set of rules as universal, raises moral solicitude and governance questions about neutrality, considering the practical implementation of whose contents are suppressed, while others of a similar kind are not taken into account. This duality of platform governance by removal of content and banning accounts cannot be a solution; it alone stands as a weapon to benefit one group or another, further concerning about the emerging issue of digital rule and divide, deepening the “North-South” dispute.