The Great Sysop: Elon Musk,X,and the emergence of platform illiberalism

Abstract

This article discusses Twitter’s mutation into X under Elon Musk, analyzing its shift from a mainstream platform to a far-right-aligned space. Through a novel conceptualization of institutional change in Trust and Safety, and an analysis of a dataset of over 1500 events related to X’s organizational transformation, we argue that three processes came to characterize X’s approach to content moderation: the political simplification of Twitter’s governance ecosystem, the centralization of power in Musk’s hands, and the repurposing of existing governance mechanisms to enforce Musk’s personal ideology. Together, these processes instantiated what we call platform illiberalism, an emerging regime whereby illiberal-esque logics reshape speech control internally while supporting illiberal actors externally. As such, X represents a unique fusion of social media and authoritarianism, with close ties to and potential implications for democratic erosion in the United States and beyond.

Introduction

“Since its launch in 2006, Twitter has turned from a niche service to a mass phenomenon” (Weller et al., 2014, xxix). So begins Twitter and Society, an influential 2014 edited volume compiling a wide range of academic perspectives on the popular microblogging service which, while never quite matching the size and global reach of other social networks like Instagram or Facebook, nonetheless came to take on an important role as a channel for news, commentary, and conversations among elites in Europe and North America (Dagoula, 2019). A space where journalists, writers, politicians, and academics occasionally rubbed shoulders with adult performers, comedians, and ordinary participants in thriving identity- and interest-based publics (Clark, 2024), Twitter—at least in the United States—earned a reputation as a vibrant left-leaning liberal talking shop.

All of this changed in 2022 when Elon Musk acquired the company. Advertisers and users left in droves as Musk revamped the service, pushing right-wing content into millions of feeds. Staffing cuts raised concerns about the platform’s continued integrity, coinciding with drastic changes to company practices that culminated in Musk’s overt support for authoritarian politicians like Donald Trump in the U.S. and Alice Weidel in Germany. Less than a year into his tenure as the “Chief Twit,” leading Internet writers were suggesting that the service—clunkily rebranded as “X”—had become a far-right platform (Warzel, 2023).

How precisely did this happen? A growing body of scholarship has emerged in recent years to explore the emergence of “alt tech” platforms—like Parler, Truth Social, and Gab—explicitly created and marketed as safe havens for right-wing extremism (Jasser et al., 2023). These services are distinctive in their claim to operate with little to no content governance mechanisms (Otero and Scharlach, 2025), thus standing in contrast to mainstream user-generated content platforms that have, for reasons of regulatory compliance and public pressure, established extensive trust and safety (T&S) bureaucracies that develop and enforce rules about acceptable behaviors online (Gillespie, 2018; Gorwa, 2019; Klonick, 2017).

Twitter’s metamorphosis into X, however, offers the first example of a mainstream global platform’s transformation into something ideologically alt-tech adjacent, and yet institutionally unique. X is now a globally active, large-user base platform that has more extensive trust and safety architecture than niche alt-tech platforms do and yet seems to wield those systems strategically for specific political aims. We identify in this transformation the emergence of perhaps the first systematic illiberal trust and safety regime in a large Western social network, which we will simply call platform illiberalism.

This article seeks to trace the emergence of this regime and understand its central mechanisms. Theoretically, our work relies on a novel framework, which defines trust and safety structures as institutions whose changes and varieties can be studied by paying attention to how rules, practices, and actors differ across three macro categories: complexity, hierarchy, and ideology. Empirically, we built a dataset of 1534 events connected with Musk’s takeover and rule until March 2025, providing a necessarily incomplete and yet illustrative record of how he transformed Twitter over time.

Using this framework and data, we demonstrate that Musk has acted as a globally influential, politically ambitious, and business-oriented type of “sysop,” the individual system operators who, as Schneider (2024) reminds us, despotically governed early online spaces. Unlike those digital rulers, however, X’s “Great Sysop” framed his despotism as rooted in the very free speech principle his actions subverted, with consequences that extend well beyond the platform.

Comparing trust and safety regimes: an institutional perspective

This section conceptualizes the trust and safety operations created by Western firms as an institution—or, more precisely, following Hall and Thelen (2009: 9), political macro institutions, made up of many components that “are themselves institutions.” Inspired by both institutionalist political economy, and the interdisciplinary literature on platform governance, we identify three conceptual variables as a starting point to understand varieties of trust and safety: complexity, hierarchy, and ideology. In each variable, we identify the features of content governance structures reflected in three concrete components: rules (normative guidelines that formally represent content moderation criteria); practices (technological, financial, and corporate procedures and actions that are dependent on unequal resources and shape the implementation of those rules), and actors, including internal and external stakeholders (see Table 1).

Table 1.

Institutional dimensions of trust and safety regimes.

	Complexity	Hierarchy	Ideology
Rules	Types of written policies	Corporate by-laws	Explicit beliefs stated in policies
Practices	Variety and nature of enforcement and policy making	Formal and informal inequalities shaping T&S procedures	Implicit beliefs embedded in enforcement and policy making
Actors	Composition of T&S teams and the role of external actors	Power concentration and tensions among T&S actors	Beliefs of key T&S actors

Institutional complexity

Complexity refers to the number, diversity, and quality of the structures involved in content governance. Research has documented how, at least on large platforms like Twitter, these structures have complexified over time, involving an ever-increasing constellation of rules, practices, and actors.

Services hosting user-generated content usually have a set of rules defining what are acceptable and unacceptable content or behaviors, as well as a method to apply them, and a system for potentially removing infringing content (Gillespie, 2018). These rules can be visible to the public, like Terms of Service and Community Standards, or internal provisions that inform day-to-day operations. Their quantity and quality vary, often according to a platform’s scale and global reach (Arun, 2022). Platforms can also adopt a myriad of other documents encompassing different aspects and phases of content governance, thus increasing normative complexity (de Keulenaar et al., 2023). This includes documents explaining how these rules are defined, enforced, and how they can be contested; policies targeting specific subjects (like copyright and advertising); or documents adhering to specific legal requirements.

The concrete governance practices related to these rules are importantly about enforcement; the human and non-human actions of identifying, analyzing, and controlling content, behavior, accounts, and personal data. But institutional complexity is contingent on decisions that go beyond the acceptability of certain content, also involving the design of rule-making procedures and internal trust and safety teams as well as the financial and technological resources they employ. These practices have become more formal and expansive. From ad hoc processes led by individuals or small teams, companies have created different policy teams and internal fora dedicated to the task (Caplan, 2018). Take, for example, Facebook’s Safety Advisory Board, or its Product Policy Forum, where employees would bring forth proposed rule changes after internal and potentially external consultation with experts and civil society. Much effort was also put into bureaucratizing manual enforcement, while technical tooling was amplified. This included expanding the use of content hash-matching systems used across different areas (e.g. for child abuse imagery, copyright violations, terrorist content), and the development of predictive automated content detection systems (Gorwa et al., 2020).

Through these processes, the number of actors involved in trust and safety has greatly increased and specialized. Beside a plethora of internal actors, large platforms increasingly depend on outsourced and precarious workers who act as content moderators (Roberts, 2018) and data annotators (Miceli and Posada, 2022). Moreover, policymaking and implementation have come to rely on consultation with civil society and other stakeholder groups (Dvoskin, 2022), which, in the case of Meta, have been formalized in a para-platform organization with limited but ostensibly binding powers: the Oversight Board. Platforms have also created or joined cross-industry groups to cooperate on the governance of content, such as the Global Internet Forum to Counter Terrorism (GIFCT—see Gorwa, 2024). Moreover, as for-profit organizations, they must address commercial demands—such as those from advertisers, users and, often, shareholders.

In principle, the more complex an institution is, the harder it is to understand and control it—for users, regulators, researchers, and the institution itself. But institutional complexity does not necessarily tell one where the loci of power are. To do that, we turn to hierarchy.

Institutional hierarchy

Hierarchy concerns the key governors in platform ecosystems and the dynamic intersection of rules and practices that produce explicit and implicit inequities between actors.

On early platforms, “sysops” physically hosted the social network’s servers in their homes and maintained authority over what kinds of posts were acceptable (Schneider, 2024). As platform governance evolved, however, the world’s most widely used services shifted power to a professional trust and safety bureaucracy subordinated to executives, boards, and founders—who, in turn, can be influenced by markets, state regulators and other platforms (such as app stores or infrastructure providers).

By-laws produce hierarchies within these bureaucracies, defining the chain-of-command embedded within corporate roles and divisions. Once in place, these ranks largely determine the procedures formalizing how content moderation decisions are made. Most of these private rules remain unknown to researchers, yet the financial structures of platforms which underwent IPOs (Initial Public Offerings) show how they can differ between companies. For instance, Meta’s financial structure ultimately gives Mark Zuckerberg the power to have a final say in all business decisions, and thus an outsize role in content moderation whenever he wishes. On the other hand, Twitter had a less concentrated, more traditional structure, in which the CEO responded to a board—until Musk acquired the company, at least (Leerssen, 2025).

In large platforms, content governance practices are expected to reflect the hierarchy embedded in these by-laws. However, many decisions influencing who ultimately has a say in content governance are made in practice rather than through formal rules. It is therefore crucial to examine how the unequal distribution of resources—financial (among teams and executives), technological (who develops and approves in-house and outsourced systems), and social (one’s embeddedness within and beyond the organization and the nature of these ties)—influences specific governance actions. For instance, different branches of platform companies might have differing capacity and autonomy to respond to policy problems, due to a distribution of resources that reflects global imbalances of power.

These somewhat open-ended relations suggest that the very creation of large content governance bureaucracies materializes as a form of power devolution—from a company’s owners and controllers to their staff and, sometimes, to external actors—through decentralization of policy making and implementation. In the latter case, the binding character of external recommendations (such as those from Meta’s Oversight Board) may play a restricted but concrete role in shaping the hierarchical configuration of content governance. Similar processes of partial decentralization are being mandated under regulatory frameworks like the European Union’s DSA (Digital Services Act), which requires that platforms allow, under certain conditions, for their content decisions to be reviewed by an external dispute-settlement body (Kira, 2025). While, in theory, these arrangements addressed a long-standing claim to involve diverse perspectives in platform policies, Caplan (2023) highlights how this “networked governance” distributes responsibilities without necessarily decentralizing power over crucial content governance decisions, potentially hampering the realization of accountability.

Institutional hierarchies are neither static nor absolute. They may wax and wane over time, and particularly salient moments (e.g. public relations scandals) may prompt the platform leadership to step in and depart from ordinary procedure. In public firms led by a board, as was the case with Twitter before Musk’s takeover, executives’ individual capacity to interfere in content governance decisions is modulated by the views and interests of directors and other shareholder representatives, as well as the decision-making capacity of the individuals entrusted to lead the company’s trust and safety apparatus. Organizations that either are not public or whose legal structure concentrates financial and corporate power in the hands of founders or CEOs are susceptible to the whims of these single individuals—which might make these firms’ regimes more unstable. Still, platforms have long claimed to incentivize internal participation and dissent. These assertions might be insincere but are not inconsequential. Leaders are routinely interrogated at companies’ town halls and employees have pushed back against rules and practices (Flyverbom, 2019).

Institutional ideology

By ideology we mean the political beliefs companies perform in public, and the ones they realize in their concrete engagement with speech governance.

Smaller platforms might have associated themselves with more radical views—including alt-right ones, which are ideologically slanted toward the needs of the far-right and supposedly “unfettered” free expression (Siapera, 2023). Yet, in large Western platforms such as pre-Musk Twitter, the policies framing content governance apparatuses evolved to reflect demands that firms align their moderation rules and practices to an idea of “international human rights” (Kaye, 2018). Beyond the centrality of freedom of expression, equality, dignity, safety, and intellectual property are widely and explicitly cited objectives of content governance structures (Celeste et al., 2023). Similarly, these companies’ discourses have often emulated liberal-democratic values, either by imprinting a constitutional language to their terms of service (Celeste, 2018) and governance initiatives (Cowls et al., 2022) or by promoting initiatives grounded on transparency, due process, and other principles of the rule of law (Suzor, 2019).

Yet their practices implicitly reveal a much more ambiguous ideological reality, underpinned by progressive neoliberalism, with universal liberal values at least on a discursive surface (Hoffman et al., 2018) and corporate commercial interests as ultimate drivers of governance configurations. In many ways, platform content moderation has realized the beliefs stated in platforms’ rules—companies have somehow sought to combat racism, hate speech, and abuse while trying to safeguard free speech (Douek, 2021). However, in their unending quest to grow their user base and avoid profit-harming regulation, they have often jettisoned social minorities’ ability to participate in the public debate (Appelman, 2025). Consider, for instance, the “mistakes” of algorithmic speech governance (with commonly used automated text classifiers exhibiting far worse performance on “low resource” languages and vernacular used by minoritized groups—Shahid et al., 2025), “colorblind policies” that decontextualize speech from “the social, political and economic advantages attendant to whiteness,” to streamline manual moderation (Diaz, 2023: 1934), and, in many cases, companies’ embrace of authoritarian governments’ demands to unfairly silence critics (Wynn-Williams, 2025).

Together, these practices demonstrate how, in implementing their business goals, platforms often reinforced oppressive socio-political structures (York, 2022), contradicting the democratic promises made in their explicit rules. Yet rather than being the result of a deliberate project aimed at promoting certain political parties, groups, and movements, ¹ these harms have arguably stemmed from companies’ (conscious and avoidable) decisions to prioritize profit above all else (cf. Magalhães and Couldry, 2025). Their approach did not seem to reflect, systematically, the political interests of any specific political actor but the assumption that a cynically selective but still non-partisan performance and enforcement of liberal values was necessary for companies to maximize the sort of openness and diversity without which their business model—entirely premised on scale—could not thrive. As we explain below, this has changed with X.

Methods

We created a dataset of 1534 public events (i.e. relevant announcements, decisions, actions, and statements) associated with Musk’s takeover and transformation of Twitter. Listed events range from a December 2017 post where Musk first mentioned in public the possibility of buying the platform to the acquisition of X by xAI in March 2025. Our factual reconstruction relied chiefly on research and news media reports in English on Twitter, X and Musk, which we collected mainly through the systematic reading of news aggregators and websites specialized in tech and tech-governance news and analysis, such as TechDirt, TechMeme, and TechCrunch. Several of these reports led to U.S. legal and regulatory filings, which we then examined. We also analyzed the company’s corporate social media accounts and Musk’s posts. For the latter, we focused on posts covered in journalistic reports and used the platform’s search functionality to identify posts addressing specific topics (e.g. Community Notes). To identify changes in policies, we used the Wayback Machine and triangulated our findings with previous work by Kopps (2024) and de Keulenaar and Kisjes (2025). Once organized in a spreadsheet, events were manually categorized by the first author using the three macro categories of our conceptual framework (complexity, hierarchy, and ideology). Rather than offering an exhaustive account, this research constitutes an initial effort to reconstruct a moment of profound change within an opaque and complex private organization. Future work may corroborate, refine, or potentially challenge elements of our analysis by uncovering new evidence—through, for example, interviews or novel documentation—or by reinterpreting facts that are already known. For a detailed methodological note and the dataset, see the online supplementary files 1 and 2, respectively.

The creation of X

Our analysis reveals a set of interlocking patterns corresponding to each category of our conceptual framework. First, we argue that Musk pursued a political simplification of Twitter’s governance regime, mainly by attacking and dismissing potentially resistant actors, inside and outside the company. This dismantling of a previously complex ecosystem facilitated a centralization of power, enabling Musk to override any other governance structure. He then wielded this power not to enforce a principled philosophy, but to advance the belief that free speech should primarily serve his own political-cum-business views and interests—resulting in the personalization of the platform’s ideology. In the following sections, events are referenced with their corresponding numbers in our dataset.

The political simplification of Twitter

While X’s public-facing moderation rules did not become less (or more) complex, Musk rapidly undermined the number and diversity of Twitter’s governance actors—promoting a political simplification of the platform’s trust and safety apparatus.

This simplification was largely executed internally. It began with his decision to take Twitter private, as he seemingly assumed that public shareholders would obstruct his political goals (29). He then acted to weaken Twitter’s leadership and trust and safety staff. Immediately before the acquisition, Musk mocked bots, trolls, and scams on the platform (2), compared then CEO Parag Agrawal to Joseph Stalin (3), and questioned Twitter’s algorithms (9), free speech record (10; 12), viability (30), and verification system (33). He continued delegitimizing the company after making an offer to buy it (43) and, afterwards, used alleged bot issues to try to exit the deal (95; 103; 112; 120). Once forced to complete the purchase in October 2022, Musk intensified his attacks, targeting Twitter’s “woke” trust and safety staff (212), orchestrating the Twitter Files scandal to expose alleged “censorship” (238; 241; 251), and falsely linking the former trust and safety chief to pedophilia (256)—leading to death threats and his forced relocation (258). As CEO, Musk rushed to slash the platform’s workforce (165). By November 2022, Twitter had laid off 80% of its 5500 employees (185). Dissent was punished: several staff were fired for criticizing Musk (188; 191). In an email, he demanded employees to either commit to a “hardcore” vision or leave (190)—prompting a mass resignation (201). Layoffs triggered lawsuits (159) and multiple tech failures (e.g. 199). Though some roles were later re-hired, reductions stuck. By September 2023, X’s full-time trust and safety team had reportedly shrunk from 230 to about 20 (687). In November of that year, the EU said that X had less than one-seventh of YouTube’s moderators (786). Fact-checking was quickly replaced with Community Notes (166), a crowd-sourced but problematic system Musk called a “game changer” for accuracy (179).

Civil society engagement also deteriorated. Musk promised a “content moderation council with widely diverse viewpoints” and that no account reinstatements would be made without its input (138)—but this council never appeared, and the previous one was disbanded after Musk accused members of supporting “child exploitation” (250). When civil rights groups urged advertisers to boycott X, Musk accused them of “destroying free speech” (161). In 2024, X withdrew from the leadership of the cross-industry GIFCT, reportedly due to a conflict over financial contributions (1256)—though this decision appears to have been reversed. Researchers were also pushed out. In early 2023, Musk ended free access to X’s API (356), hampering external study of the platform. Critical organizations documenting governance changes faced attacks and lawsuits (606; 681; 720; 804; 811; 850). Still, X continued to claim to collaborate with select civil society partners focused on child sexual abuse material (CSAM; 877) and synthetic drug content (1121)—issues especially aligned with conservative moral priorities in the United States.

Musk took perhaps the most decisive action against advertisers—while still courting them as business partners. In April 2022, he tweeted that Twitter should have “no ads,” saying corporate influence over policy was stronger when platforms relied on ad revenue (28). This marked the start of his efforts to reduce Twitter’s ad dependency by pushing a subscription model that effectively monetized Twitter’s verification system. Framed as a form of visibility democratization (150), this change subverted the platform’s former elite of verified users and eventually transformed X into a “freemium” platform (e.g. 361; 414; 427; 489). Musk also pushed to make X an “everything app” like WeChat (654), promising to enable payments and other financial services (846; 891; 1104; 1464) and began using users’ data to train products of his AI company, xAI (1281). Yet there’s no evidence that new revenue streams were enough to replace advertising money. X tried to win companies back by recreating its Client Council (623) and launching technical measures: in December 2022, it introduced “adjacency controls,” to prevent ads from appearing near controversial content (270) and expanded these in 2023 and 2024 (346; 619; 718; 929), leading to a brand safety certification (1098). Advertisers were not convinced (67), and beginning in November 2022 (210), largely abandoned X (e.g. 240; 307; 757), likely due to diffuse reputational concerns and falling user numbers (e.g. 996). Musk’s reaction ranged from trying to appease them (135) to aggressive verbal and legal retaliation (838; 1153), with the help of far-right actors in Congress and Republican State Attorney Generals (1342; 1120).

Generally, a hallmark of Musk’s speech governance practices was his attempt to replace human moderation with automation, making the platform more “engineering-driven” (190). This was built into the “freedom of speech, not reach” policy (more on this shortly), where posts’ visibility had to be algorithmically controlled and evaluated against moderation policies. Allegedly, new automated systems were used for moderating CSAM (1134) and enforcing a narrower definition of hate speech (418). It is difficult, however, to independently assess these claims—there is evidence that the hash-matching for known CSAM was disabled in June 2023, for instance (Woollacott, 2023). Even if they were indeed expanded, such systems may have furthered political simplification as their opacity made them harder to be challenged.

The centralization of power in Musk’s hands

The consequences of Twitter’s simplification were not preordained. Musk could have restaffed the company with people he trusted, allowing them to independently rebuild the platform’s content moderation regime. Instead, X became acutely centralized around Musk himself.

This process involved the institution of unprecedented practices of algorithmic rigging. There is evidence that, in 2024, his engineers twice acted to prevent Grok, xAI’s LLM embedded on X, from producing content critical of Musk (1514; 1517). Yet the most consequential rigging happened in February 2023, when Musk, reportedly unsatisfied with the reach of his posts, ordered engineers to redesign Twitter’s recommendation system to artificially boost his own visibility by a “factor of 1000” (379), something he publicly acknowledged with a meme (382). Given his changes to the platform’s API, the amplification of Musk’s visibility remained difficult to study. Yet public evidence from 2024 suggests that his account continued to be artificially boosted (1336), which, combined with his frantic posting habits and increasing political relevance, has caused his visibility on the platform to skyrocket. In March 2023, Musk became the platform’s largest account with 133 million followers; by November 2024, this number had increased by over 50% to 202 million (1310); in March 2025, he had 218 million followers. In engineering his quasi-omnipresence, Musk transferred central aspects of the platform’s speech governance system to his individual account, overriding previous organizational procedures. Whatever he decided to say, share, or engage with was likely to be shown to whoever stayed long enough on the platform. In this way, mass content amplification—previously a multi-actor process shaped by established rules and practices—has come to importantly hinge on the whims of a single person.

Despite that early acknowledgment of his rigging of the recommendation system, Musk never fully admitted that X’s news feeds were redesigned to benefit him. Trying to prove he did not enjoy any special treatment on the platform, he said several times that his posts were susceptible to Community Notes (e.g. 593; 951). However, not only have a tiny proportion of Musk’s misleading posts been Noted (1255), but the artificial amplification of his account could cancel out the negative effects that fact-checking labels had on a post’s circulation. Furthermore, despite Community Notes’ promise to de-concentrate content moderation by empowering ordinary users, X (and Musk) retained unilateral control over the mechanism that allows a Note to appear.

The steep centralization of power under Musk can also be seen in how the platform’s formal rules have been disrupted and even made irrelevant by his individual interventions. In December 2022, the platform introduced a new policy prohibiting the sharing of someone’s live location after Musk disagreed with a user who tracked his private jet (268; 272). In February 2023, the platform changed its “graphic violence” policy after Musk criticized the flagging of a Republican Senator’s picture (370). Another policy that Musk unilaterally changed regards transgender individuals. In April 2023, the platform dropped protections against deadnaming and misgendering after Musk said that his own posts could violate this policy (454). Then, in early 2024, the protections were partially reinstated in diluted form (908). In response, anti-trans accounts complained directly to Musk, who replied that the reintroduction of protections “was due to a court judgment in Brazil” and “should not apply outside of” the country (970); the policy was then updated to add the sentence that the provision would only apply when “required by local laws” (1030). On at least one occasion, his wishes did not need to become a public rule to be enforced. In June 2023, he posted that the term “cisgender” would be considered a slur on the platform (533); in May 2024, users started reporting that they had been punished for using it—a prohibition that appeared nowhere in X’s published policies (1072). This is not to say that all of his public demands materialized into new rules. For instance, Musk said that the platform would delete the block button (515)—which was watered-down but did not disappear (1275), likely because app stores demand or suggest apps have this functionality. He also said that users whose posts were repeatedly Noted should be somehow penalized (441)—something that never occurred, perhaps because he has, over time, been Noted at least 127 times (Community Notes Leaderboard, 2025), even if this corresponds to only a fraction of his problematic posts (as explained above).

The individualization of Twitter’s ideology

Musk could have used his centralized powers to enforce any ideology. However, the only consistent principle behind his view of free speech seems to be that content moderation should primarily serve his own interests and views, which have aligned with those of the global far-right.

Musk has repeatedly framed X as a platform centered on free speech, not profit. Since before the acquisition, he called himself a “free speech absolutist” (7), a vision whose three pillars he defined as: (1) speech should be “as free as possible,” (2) the deletion of content or banning of users should be avoided, and (3) safeguarding free speech means protecting “the future of civilization,” and as such is a moral not economic goal (38). Once CEO, he announced a new policy: “freedom of speech, but not freedom of reach,” meaning content might be “deboosted & demonetized” rather than deleted (196). The term “free speech” became X’s motto (840; 1001; 1084), and even critics commonly described the platform as a libertarian space (Davies, 2024).

Our analysis challenges this assumption. Take first the contradictory shifts in content moderation rules. Almost all policies that were relaxed addressed, in fact, grievances of the far-right: removing COVID-19 guidelines (214), weakening “Harassment and Abuse” rules (385), seemingly eliminating the “Coordinated Harmful Activity” policy (440), and dropping provisions about electoral results disinformation, of particular importance for Trump supporters (650). Other rules, in fact, expanded control, still echoing the same actors’ concerns and discourse—like broadening policies on “doxxing” (995) after the real name of a pseudonymous far-right cartoonist spread on the platform (987), and strengthening child safety policies (1074). Meanwhile, the “X Rules” header continued to emphasize the need to balance freedom and safety (X, 2025), a traditional discourse among large platforms that X never fully abandoned (794; 918; 1262; 1513).

The selective nature of Musk’s free speech stance was even clearer in X’s content moderation practices. Musk used his self-given visibility to amplify falsehoods and hate, becoming by late 2024 a “superspreader” of misinformation and racist narratives (1249; 1291; 1292). He heavily promoted authoritarian actors including Donald Trump (1177), AfD leader and German far-rightist Alice Weidel (1434), British agitator Tommy Robinson (1413), and anonymous extremists (e.g. 1531). He also intervened to protect allies and undermine critics. A key move was granting amnesty to most suspended accounts after polling users in November 2022 (213)—nominally universal but largely benefiting the right (see Mosleh et al., 2024). At the same time, staff were reportedly blocked from enforcing rules against the far-right Brazilian President Jair Bolsonaro in his 2022 failed reelection attempt (1070), and during the 2024 U.S. elections, X blocked the circulation of a dossier on then VP candidate J.D. Vance (1250). Journalists critical of Musk were penalized multiple times, including suspensions related to said private jet tracker (275; 276), the suppression of investigative outlet Bellingcat after Musk’s critique of it in 2023 (480; 483), and the cancelation of Don Lemon’s X show in 2024 following a critical interview with Musk (984). Activists also faced consequences, such as Irish anti-far-right accounts punished in late 2023 (832), and posts about DOGE employees deleted in 2025 (1476), all under the pretext of doxxing. Even MAGA activists accused Musk of censorship after disagreeing with him on immigration policies (1404).

Similarly, X’s dealings with governments suggested not a principled defense of free speech (1248) but the use of the platform to promote Musk’s business interests and political sympathies for authoritarian actors. Neither he nor the platform resisted or complained about content takedown requests from China (237), which is central for Tesla’s sales and production (502; 1487). In Brazil, however, after the electoral loss of Bolsonaro, Musk clashed with Supreme Court Justice and Bolsonaro’s critic, Alexandre de Moraes, and refused to obey judicial orders against Bolsonaro’s supporters (1239; see 88; 243). Conversely, Musk maintained a warmer relationship with India’s ethno-nationalist Narendra Modi, of whom Musk said he was a “fan” (534). In August 2022, he opposed Twitter’s pre-existing legal challenge to India’s content regulation (115). After the acquisition, X removed anti-Modi content repeatedly (342; 428; 446), although some disputes occurred (957; 1525).

X never explicitly admitted its alignment with authoritarianism. Rather, there was a constant attempt to portray the platform as democratic and representative of its users. This was made obvious by Musk’s use of the platform’s poll feature to justify controversial moves, often followed by the populist expression “vox populi, vox dei” (203; 213; 290; 291; 853; 1139). But the polls were likely unrepresentative (305) and not always consequential—he stepped down from X’s CEO position after a poll he launched suggested that most users wished him to do so, for instance, but still retained control over trust and safety decisions (604). X corporate accounts tried to position the company on the side of “the people” and not of the “elites,” also adopting a populist tone (842; 1326; 1358; 1364; 1431). The Community Notes program played an important role in this discourse, with Musk touting its bottom-up nature (231). Yet, the program could be organically gamed by users blocking Notes from appearing (289; 545; 564).

This individualized ideology damaged X’s reputation and user experience, undermining the company’s performance—until it rebounded, mainly for political reasons. Despite Musk’s initial claims, his actions showed clear business intentions, including his attempt to back out of the deal for financial reasons (146) and, afterwards, efforts to rebuild the platform’s business model. But ad revenue fell drastically and wasn’t offset by subscription growth (232; 511; 707). By September 2024, X’s value was marked down by 79% (1254). After Trump’s 2024 win, however, the company’s valuation returned to that original level (1530), when Musk sold it to his AI company, xAI, for US$45 billion (1534). This reversal likely stemmed from investors’ confidence that Musk would leverage his influence over Trump to benefit his own businesses (1395). Soon after, major advertisers like Amazon, Disney, and Apple returned—likely fearing political backlash (1332; 1470; 1490) and responding to explicit threats (1503).

Platform illiberalism

Drawing upon conceptual models from political science, particularly from work that has tried to make sense of current forms of authoritarianism, we argue that the creation of X instantiated a form of illiberal speech governance. Platform illiberalism, as we describe it, entails two entangled processes: the deployment of ambiguous illiberal-esque logics to govern speech within the platform, and the strategic mobilization of this regime to support illiberal actors operating beyond the platform.

Illiberalism has been used to refer to political processes that lead to the visible breakdown of the norms of liberal democracy in different nation states—echoing similar notions of democratic “erosion,” “backsliding,” or “decay” (Bermeo, 2016; Lindstaedt, 2021). While the precise contours of these concepts vary, they broadly refer to “the interplay of authoritarian-minded political leaders and their project to transform an existing democracy into an alternative system of rule” (Kneuer, 2023: 845), in particular through the deprivation of basic rights and freedoms (Zakaria, 1997: 22). Illiberalism thus can be applied to a range of scenarios, from sharp processes ultimately leading to regime change to slower transformations that decrease the quality of political institutions (Iglesias Keller and Arguelhes, 2024: 189).

It would be inaccurate to directly transpose insights on state illiberalism to the considerably different reality of a platform—which, as a private company, has no elections, parties, army, or real courts. Yet, once we understand illiberalism as a general governance logic that “undermines liberal democratic institutions and values by ostensibly leveraging the very frameworks it seeks to challenge” (Štětka and Mihelj, 2024: 26–27), it becomes clearer how this concept helps us make sense of the institutional transformation that took place on Twitter.

Consider first the three dynamics we have outlined. Musk’s political simplification of the platform governance ecosystem is comparable to what has happened in states that saw the “debilitation or elimination of the political institutions sustaining an existing democracy” (Bermeo, 2016: 5). Illiberal actors often dismantle potential accountability mechanisms that could apply checks to their power—such as public administrators, judicial bodies, and independent media. Power centralization is thus typical of these scenarios. It may include, for instance, the capture of institutions designed to secure free and fair elections, which often happens through, or in parallel with, executive aggrandizement, whereby elected officials undertake “a series of institutional changes that undermine the capacity of opposition forces to contest executive preferences and secure political power” (Bermeo, 2016: 11). The ideological parallels are also striking. In adopting and promoting the bigoted views of the far-right, X’s content governance has enforced creeds that largely correspond with those of state illiberalism, especially the rejection of multiculturalism and minorities (Laruelle, 2022: 307–309), and their belittlement through hateful, conspiratorial, and misleading speech.

Another similarity involves how the platform deployed misdirection, ambiguity, and contradiction. In the same way that illiberal politicians capture existing institutions to destroy liberal democracy, Twitter’s governance structures were repurposed to weaken freedom of speech in its own name. Content moderation was not systematically diminished, much less abolished. Instead, platform speech control was redirected to realize Musk’s individual views and goals. Some structures—in particular Twitter’s reliance on civil society organizations and fact-checking—were heavily degraded. But even then, the platform did not discursively abandon cooperation with external actors or the possibility of content labeling, for instance. Central to this ambiguity was the relentless attempt to frame Musk’s reign as a principled, historical fight against “woke” censorship, the use of polls and Community Notes to give the impression that users could have a final say in major moderation decisions, and the platform’s defensiveness against those who criticized its trust and safety system. Finally, in the same way that illiberal governments can tolerate some criticism, the platform does not enforce Musk’s ideology in a totalitarian manner—millions of users routinely criticize him, his allies, and their ideas. Yet not only does such freedom of speech rarely translate into freedom of reach: he has also shown that he will intervene to shape participation on the platform when and if necessary, exerting a sort of potential tyranny akin to that wielded by the far-right sysops of forums like Kiwi Farms and Stormfront (Törnberg and Törnberg, 2025).

Platform illiberalism involves more than the transplantation of illiberal logics of power to content governance: it needs to be considered against a wider global trend of democratic erosion—specifically how its approach to trust and safety relates to illiberal movements and state actors.

Most evidently, this relation concerns platforms’ communicational power—a key component of how economic elites, including tech leaders, have tried to retain influence “in the face of disruption and crisis” (Griffin, 2024: 4). As our analysis demonstrates, the creation of X was an indisputable attempt to turn a mainstream platform into a propaganda machine that supports authoritarian projects in the United States, Germany, and the United Kingdom, among others. In line with the complex relationship between digital platforms and the turn toward the right in various global elections (Lorenz-Spreen et al., 2023), the impact of this machine remains unclear. Yet, even in the case that Musk’s efforts have not decisively changed voting results, they have clearly expanded the institutional horizon of illiberalism. Democratic erosion often comes with the expansion of “illiberal public spheres,” the “communicative space comprising both traditional and new media that promote and amplify illiberal actors, views, and attitudes” (Štětka and Mihelj, 2024: 29). X has shown that the social media dimension of illiberal communication need not be limited to small, fringe alt-tech platforms: even a large, somewhat liberal company can be quickly repurposed, co-shaping what people (including political elites) pay attention to, know, and believe. And these issues are not limited to traditional forms of political messaging. Platform illiberalism inevitably transformed the X data used to train Grok—influencing how the LLM privately “speaks” with users and responds to prompts on a range of topics (for a related discussion, see DiResta, 2025).

Platform illiberalism also contributes to a broader undemocratic political economy where private and government interests increasingly collude. One can argue that the overt politicization of X, together with Musk’s substantial financial contributions during the 2024 U.S. elections, positioned him as a major political actor and ultimately enabled an unusually high degree of power concentration in him during the initial months of the second Trump administration. That juncture proved fleeting. But it epitomized the rise of a new oligarchical force, which the breakup between the tycoon and the President hardly stopped (Cohen, 2025). In blazing the trail followed by other tech leaders who have since kowtowed to Trump, Musk’s political success with X made attractive the idea that high-tech capitalism can only fully thrive when unconstrained by the rule of law and, perhaps, democracy itself. Importantly, the extraordinary individual power he gained and exerted on and due to X can only be understood in relation to long-standing structures, such as regulatory and corporate frameworks that enable wealth accumulation and market consolidation while overlooking the political influence these processes create.

The emergence of platform illiberalism clearly entailed a shift from Twitter’s own origins—marked by bottom-up dynamics, whereby the company would adopt and embed users’ tools and views (Burgess and Baym, 2020). But can X’s example be a blueprint for how other large Western platforms transform vis-à-vis renewed authoritarian pressures? In some ways, the process this article examined appears unique, driven by Musk’s particularly erratic leadership style and extreme views. Furthermore, how digital platforms configure their operations is contingent on ever-changing legal and corporate factors, which might differ from those that shaped the creation of X. That said, after the election of Trump in 2024, other companies have at least partially followed in X’s footsteps. Meta has made public-facing changes to its policies on issues at the heart of the global far-right’s culture war agenda and trimmed back its collaborations with external fact-checkers in the United States, to some extent replacing them with X-inspired Community Notes (Zuckerberg, 2025). YouTube tested a similar system and relaxed policies traditionally disliked by conservatives (Grant and Mickle, 2025). These changes did not necessarily follow the internal illiberal logics of X but they did cater to the illiberal political project led by the Trump administration. The acquisition of TikTok’s operation in the United States by Trump allies might prove to be yet another example of authoritarianism shaping a large social media platform (McMahon, 2025). As these changes unfold, it is possible that platform illiberalism will become multiple—occurring in different ways, and with varying intensities, in distinct organizations.

Documenting, analyzing, and confronting these developments should be at the top of the agenda of researchers, citizens, and policymakers alike. To some extent, the resistance has already begun. Musk’s political role caused a global consumer backlash against his main company, Tesla (Milmo and Jolly, 2025), contributing to his distancing from the U.S. government. Furthermore, the European Commission was investigating whether X violated the DSA and has fined the company €120 million for breaching transparency obligations (European Commission, 2025). Ultimately, whether platform illiberalism will be contained depends on our collective capacity to see tech companies’ ingratiation with authoritarianism for what it is: a strategy that is neither natural nor necessary but a political decision that further undermines the possibility of a democratic social order. We hope to have contributed to this denaturalization.

Supplemental Material

sj-docx-1-nms-10.1177_14614448261424889 – Supplemental material for The Great Sysop: Elon Musk, X, and the emergence of platform illiberalism

Supplemental material, sj-docx-1-nms-10.1177_14614448261424889 for The Great Sysop: Elon Musk, X, and the emergence of platform illiberalism by João C. Magalhães, Clara Iglesias Keller and Robert Gorwa in New Media & Society

Supplemental Material

sj-xlsx-2-nms-10.1177_14614448261424889 – Supplemental material for The Great Sysop: Elon Musk, X, and the emergence of platform illiberalism

Supplemental material, sj-xlsx-2-nms-10.1177_14614448261424889 for The Great Sysop: Elon Musk, X, and the emergence of platform illiberalism by João C. Magalhães, Clara Iglesias Keller and Robert Gorwa in New Media & Society