Whack-a-Mole No More: How Extremists Choose Their Digital Safe Havens

Tamar Mitts’s “Safe Havens for Hate” arrives at a moment when debates over online extremism and platform regulation are increasingly urgent—and increasingly polarized. Governments have begun to implement more stringent rules on the largest social media companies, while extremist movements continue to migrate and adapt, exploiting gaps in the regulatory landscape. Mitts seeks to explain this dynamic, and she does so with an ambition that distinguishes her work from much of the more anecdotal or narrowly focused literature in the field: She draws on data tracking the online activity of 114 militant and hate organizations across 67 platforms, supplementing this with qualitative case studies of the Islamic State, the Proud Boys, QAnon, and the Taliban.

Mitts structures her findings around several key arguments:

1. Online platform regulation encourages uneven platform moderation. Regulators tend to focus on the largest platforms, imposing heavier obligations there. Smaller platforms, especially those below regulatory thresholds, become refuges. Telegram’s proximity to, but avoidance of, the EU’s “very large online platform” designation is one key example.
2. The uneven moderation landscape drives extremist migration. When choosing platforms, extremist groups do not simply flock to the spaces with the least moderation. Instead, they seek the best combination of low enough moderation to tolerate their content and high enough reach to maximize their impact. This produces a patterned and, Mitts argues, predictable migration across platforms.
3. Moderation disparities facilitate mobilization and recruitment. When larger platforms adopt stricter moderation policies that lead them to expel, or “de-platform,” users who violate their policies, they push resentful audiences to less-regulated spaces where they are primed for radicalization. Extremists exploit these grievances in their recruitment strategies.
4. Adaptation to moderation helps with survival and message relevance. Groups tailor messages to platform rules, avoiding explicit language or imagery on mainstream sites while pushing more explicit content elsewhere. QAnon’s appropriation of the benign-sounding #SaveTheChildren on Twitter is one illustrative tactic. Adapting to moderation helps groups evade platform enforcement while enhancing the resonance of their message with particular audiences.
5. Convergence in moderation can constrain activity—but at a cost. When multiple platforms align on stricter content rules, extremist organizations find it harder to maintain official accounts and are more likely to be banned. However, they also shift to informal dissemination via unofficial accounts.

One of the book’s strengths is its commitment to empirical grounding. Mitts moves beyond the “whack-a-mole” cliché of de-platforming debates, showing that migration is not chaotic but shaped by structural incentives. Her cross-platform analysis, rare in a literature that tends to focus on Facebook, YouTube, and X (formerly Twitter), helps to foreground overlooked sites. The case studies inject human texture into otherwise abstract findings, whether in the Taliban’s platform strategy or QAnon’s rhetorical pivots. For readers weary of theory without data, “Safe Havens for Hate” offers a welcome corrective.

Where the Argument Delivers—and Where It Falls Short

For all its contributions, “Safe Havens for Hate” also reflects the limits of its framing, methods, and scope. These do not diminish its value, but they matter for understanding how far its conclusions can be taken and where future research might push further.

Narrow Framing of Regulation

Mitts’s discussion of regulation focuses almost exclusively on content-based rules: laws or policies designed to suppress categories of “harmful” speech as defined in each regulation. This framing underplays other regulatory approaches that have gained traction in recent years, such as design-based interventions (altering platform features that facilitate extremist networking), transparency mandates (reports with aggregate data on moderation actions, information about how algorithmic systems work), and procedural requirements (appeals processes, risk assessments).

Her typology of regulatory scope also tends to oversimplify. Laws are presented as falling into two categories—applying equally to all platforms or targeting only the largest—without sustained discussion of hybrid or layered models. For example, the United Kingdom’s Online Safety Act, which casts a wide and ambitious scope in terms of types of platforms covered (that is, any site that allows users to share content or interact with each other, including online games, dating apps, and instant messaging services), is painted with the same brush as regimes such as the European Union’s Digital Services Act, which applies to a relatively narrow type of platform—those that enable dissemination of information to an unlimited number of users. The result is a somewhat flattened picture that obscures how regulators balance breadth with enforceability.

This matters because Mitts’s findings on the consequences of uneven moderation could apply beyond content rules to age verification mandates, parental controls, or data privacy standards. The tension she identifies between focusing on the biggest players, which have the most impact and greater resources for compliance, and inadvertently driving harmful activity into less-regulated spaces is a recurring challenge in tech governance, not just in extremist content regulation.

Platform Scope and Potential Blind Spots

The dataset’s breadth is one of the book’s strengths, but the definition of “social media” it implicitly adopts produces some notable exclusions (Mitts doesn’t list the 67 platforms, but her data analysis and visualizations indicate a narrower scope than this high number might suggest). Messaging services such as WhatsApp—massive in scale but often overlooked because they are perceived as private spaces—are absent. So too are gaming platforms, which combine large audiences with affordances (e.g., live voice chat among strangers) that are attractive to extremist actors. Both categories illustrate that a platform’s regulatory status, moderation threshold, and appeal for extremists are not always determined by its size.

Mitts also leaves aside the dark web, a choice she acknowledges but does not explore in depth. While her interest is rightly in the public-facing online ecosystem, omitting these other arenas limits the generality of her claims.

Methodological Simplifications

Perhaps the book’s most significant limitation is Mitts’s measure of platforms’ “moderation thresholds.” She derives these from counting the number of prohibited content categories in a platform’s terms of service or community guidelines. This assumes a neat mapping between written rules, actual enforcement, and user perceptions.

Yet as Mitts herself concedes elsewhere in the book, enforcement often diverges sharply from stated policy, and extremist groups may base their strategies on lived experience or secondhand perception rather than the letter of the rules. Some companies are transparent about lenient enforcement; others maintain elaborate, rights-respecting policies that they quietly ignore. Reducing moderation to a policy tally risks misclassifying such platforms and oversimplifying the strategic calculations of extremist actors.

Moreover, the analysis tends to equate “impact” with audience size. While reach is undeniably important, other features—private messaging, anonymity, voice chat, algorithmic recommendation systems—can also shape a platform’s value to extremist groups. Mitts notes this in her discussion of the Islamic State’s move from Twitter to Telegram, where encryption and privacy mattered alongside audience, but she does not integrate such factors into her broader conclusions.

Assumptions and Missing Counterfactuals

The trade-off between reach and moderation sits at the heart of Mitts’s framework. While compelling, this trade-off does not always exist: Platforms such as X under Elon Musk offer both substantial reach and relatively low moderation, blurring the neat correlation she identifies between size and restrictiveness. Similarly, Telegram occupies a sweet spot of large scale and limited regulation.

Her treatment of moderation convergence—the idea that aligning thresholds across platforms can constrain extremist activity—also invites questions she does not fully address. Would uniform moderation rules, applied everywhere, simply deepen resentment among anti-establishment audiences? If so, how should policymakers weigh the trade-off between reducing extremist visibility and fueling narratives of censorship?

Underdeveloped Policy Implications

After diagnosing the problem in depth, Mitts devotes just two paragraphs of her conclusion to possible responses. One recommendation—encourage convergence without restricting lawful speech—runs directly into the normative and practical challenges she herself has outlined, including the risk of “censorship creep” if anti-democratic governments influence shared moderation databases. The other—make moderation more legitimate in users’ eyes—seems aspirational at best, especially given the lack of consensus on definitions of extremism and “harmful content.”

The brevity of this section leaves the reader with a solid grasp of the dilemma but little sense of how to begin addressing it. This is not fatal—descriptive work is valuable in its own right—but it limits the book’s utility for policymakers or practitioners seeking actionable guidance.

What We Can Take Forward

Despite its limits, “Safe Havens for Hate” offers insights worth building upon. First, it reframes migration as strategic rather than chaotic. Extremist actors do not simply scatter when de-platformed; they make calculated moves based on structural differences between platforms. Recognizing this can improve both research and policy by encouraging a cross-platform rather than siloed view of online extremism.

Second, Mitts underscores that moderation disparities affect not just where extremist content appears but how it evolves. When groups tailor messages to platform rules, harmful content does not vanish, it shifts form and venue. This means regulators and researchers must track patterns across the ecosystem, not just content on individual sites.

Third, she acknowledges that centralizing moderation power, whether through regulation or industry coordination, carries risks. The “censorship creep” she warns about is real, particularly in less democratic contexts. The challenge is to close loopholes without creating new tools for repression.

For future work, Mitts’s framework could be refined in several ways:

• Measure moderation in practice, not just on paper. Incorporate enforcement data, user perceptions, and qualitative accounts from platform insiders.

• Account for affordances beyond audience size. Design features can be as important as reach in shaping a platform’s utility for extremist actors.

• Broaden the regulatory lens. Consider procedural and design-based interventions alongside content rules.

• Expand the platform universe. Include messaging services, gaming platforms, and other high-impact spaces outside conventional “social media.”

Why does this matter now? Because the correlations Mitts identifies—between platform size and moderation strictness—are already being disrupted by Big Tech retrenchment in trust and safety, shifting political winds, and the proliferation of mid-sized platforms. Uneven regulation will persist, and extremist adaptation will accelerate. Understanding these dynamics is essential to designing interventions that work in the real, messy online ecosystem.

In the end, “Safe Havens for Hate” is not the final word on extremist platform migration, nor does it pretend to be. Its strength lies in pairing an ambitious dataset with vivid case studies to illuminate a problem that is often discussed but rarely mapped in this much detail. Its shortcomings—narrow regulatory framing, methodological simplifications, and thin policy prescriptions—should be read not as fatal flaws but as invitations for the next round of scholarship. If Mitts’s goal was to move the conversation beyond metaphors and toward patterns we can actually measure, she has succeeded. The next challenge is to turn those measurements into sound policy strategies.