Mona Shtaya in collaboration with our partners Beam Reports and Kashif.
Background
Tech companies play an increasingly central role in shaping access to information during conflicts. This brief digs into the implications of content moderation practices during crises, particularly in the SWANA region.
During armed conflicts, the ability to access timely, accurate, and life-saving information is critical. Yet in many parts of the Global Majority, especially in conflict-affected regions like South West Asia and North Africa (SWANA), social media platforms are failing to meet this need. The recent rollbacks—alongside weaker hate speech policy—have undermined access to credible information and enabled the spread of disinformation. These changes are not only silencing marginalized voices, but also putting lives at risk. This brief discusses the consequences of such shifts and calls on tech companies and policymakers to adopt accountable, rights-based moderation policies and practices that ensure safety and information access during crises.
Although tech companies often serve as primary channels for communication, eyewitness testimony, and real-time documentation of human rights abuses, the spread of disinformation—particularly in countries experiencing authoritarianism or conflict—continues to have a devastating impact on people’s lives. It undermines their ability to make informed decisions, endangers their lives, polarizes public opinion, and can contribute to radicalization and extremism. This brings up the question: to what extent were the recently made policy changes informed by meaningful consultation and proper human rights due diligence?
Tech companies’ growing influence has come with significant responsibility, especially in conflict-affected regions such as SWANA. Meta, like many tech companies, has severely neglected its responsibilities in some of the world’s most violent conflict zones. Internal documents reveal that despite high risks, the company invested minimal resources to tackle harmful content.
Tech companies’ policies and enforcement mechanisms—largely shaped by tech companies based in the Global Minority—often fail to reflect the needs and realities of users in the Global Majority. In essence, while content moderation practices are intended to curb disinformation and hate speech, they often result in over-censorship that disproportionately affects marginalized voices. In conflict-affected areas, this overreach can lead to the deletion of content that documents war crimes, provides crucial information, or offers critical updates during communication blackouts. People living in some of these regions are already denied equal access to a broad spectrum of online content because of the digital divide. Removing credible, informative material—or failing to address harmful disinformation—can put lives at risk and hinder their ability to understand unfolding events. This dynamic is especially evident in contexts such as Sudan, Palestine, Syria, among other countries, where communities face the dual burden of conflict and digital oppression, exacerbating the challenges they endure in times of crisis.
Content Moderation in Times of Conflict: Tech companies’ content moderation policies and practices during conflict and their impact on people’s ability to access life-saving information
As social media platforms become a primary source of news and updates, particularly during crises, content moderation policies of companies like Meta, TikTok, X (formerly Twitter), and Telegram are increasingly undermining access to vital, potentially life-saving information. These policies not only limit people’s right to access information about crucial updates, but also weaken efforts toward accountability for war crimes and human rights violations.
Three main issues are contributing to this growing problem:
1. Over-Enforcement: Silencing Critical, Life-Saving Content
Social media platforms often rely on automated moderation tools that lack contextual understanding, particularly in some dialects and languages, such as Arabic, and certain local, historical and political contexts. This results in the over-enforcement of policies that remove videos, images, and eyewitness accounts—often uploaded by journalists, civilians, or civil society organizations—which are essential for public safety, documentation, and legal accountability.
For example, during Sudan‘s ongoing conflict, community-led groups used platforms like Facebook and WhatsApp to share evacuation routes, hospital locations, and safety instructions. When such posts were taken down or algorithmically suppressed, civilians were left without critical information, putting their lives at even greater risk. Similarly, in Palestine, many firsthand accounts and evidence of potential war crimes were removed or demoted under policies meant to suppress violent content, further contributing to the erasure of documented atrocities.
2. Under-Enforcement: Hateful Speech and Disinformation Flourish
At the same time, these platforms often fail to adequately enforce policies against hate speech and disinformation, particularly when it targets marginalized communities. In 2023, Insecurity Insight reported that doctors and humanitarian workers in Sudan were repeatedly targeted with misinformation, disinformation, and hate speech, which severely compromised their safety and impeded humanitarian access. An analysis of nearly 1,000 Facebook comments from April to May 2024 further confirmed that misinformation, disinformation, and hate speech remain a major threat in Sudan’s online space. Additionally, in Syria, coordinated networks on X have been used to spread sectarian disinformation in the aftermath of regime shifts, further destabilizing an already fragile information ecosystem. These disturbing narratives not only inflame tensions but also endanger humanitarian personnel’s well-being and operations in conflict zones.
On August 10, an Israeli drone strike hit Al Jazeera’s tent office inside the Al-Shifa medical complex in western Gaza, killing the entire team of four staff journalists in the city—including Anas al-Sharif—as well as two freelance reporters, one of whom was working for Al Jazeera. Prior to that, in July 2025, the Committee to Protect Journalists (CPJ) and the United Nations issued separate statements raising grave concern for the safety of Anas al-Sharif, Al Jazeera Arabic’s Gaza correspondent, following an intensified online smear campaign led by the Israeli military’s Arabic spokesperson, Avichay Adraee. Adraee’s videos, circulated on Facebook and X, repeatedly accused al-Sharif of terrorism and propaganda, including mocking footage of him crying on air while reporting on starvation in Gaza. Irene Khan, UN Special Rapporteur on freedom of expression, condemned the attacks as “unsubstantiated” and “a blatant assault on journalists.” CPJ warned that such unfounded allegations appeared designed to legitimize targeting the journalist. The attack underscored the lethal consequences of online smear campaigns, where untackled disinformation and incitement spread through social media platforms can pave the way for real-world threats against journalists.
3. Algorithmic Amplification of Misinformation and Disinformation
Beyond enforcement gaps, tech companies—particularly Meta, X, YouTube, and TikTok—play a defining role in shaping the misinformation, disinformation, and malinformation (MDM) ecosystem. Their algorithmic design prioritizes engagement over accuracy, allowing sensational or misleading content to spread more widely than factual reporting. Research shows that posts from low-credibility sources, especially when shared by influential users, often receive greater visibility than reliable journalism. This amplification contributes to echo chambers, political manipulation, and the spread of extremist narratives, with tangible offline harm ranging from public panic to hate-driven violence.
The lack of transparency and accountability compound the problem. Tech companies rarely disclose how algorithms amplify or suppress content, and voluntary transparency reports offer little clarity. This opacity is especially harmful in the SWANA region where authoritarian regimes actively weaponize disinformation while suppressing independent voices. Without accountability or independent oversight, MDM continues to erode information integrity and endanger vulnerable communities.
One concrete example comes from Sudan, where disinformation and hate speech have been weaponized against marginalized communities, fueled in part by algorithmic amplification. Beam Report’s Sudalytica documented a sharp rise in hate speech targeting Kanabi residents in Sudan’s Gezira agricultural project—mainly Fur, Zaghawa, and Hausa communities. Between November 15, 2023, and November 15, 2024, the report recorded over 3,690 hateful posts on X, reaching nearly 29.9 million views and 10.6 million engagements.
Translated text: Most of al-Kanabi people have joined the Rapid Support Forces (RSF), and Al-Gezira is full of al-Kanabi to the extent that Ethiopian snipers are sniping the army from al-Kanabi… the pictures are clear.
By mid-December 2023, online narratives began warning that Al-Kanabi residents posed a “potential threat,” with baseless claims that they could serve as a “human reservoir for militias” aligned with the Rapid Support Forces (RSF). As the conflict escalated in 2024, the rhetoric intensified, shifting from vague warnings to direct accusations that most Kanabi residents had joined the RSF. These unverified claims fueled incitement to violence, including calls for forced displacement and harm. By late October 2024, conspiracy theories circulated portraying Kanabi residents as part of an RSF-backed plot to replace Gezira’s “indigenous population”—framing them as an existential threat to be eliminated.
The amplification of this disinformation across major platforms normalized hate speech and dehumanization, directly contributing to offline harm. After the RSF’s attack on Gezira in December 2023 and during the Sudanese Armed Forces’ (SAF) advancement in late 2024, online campaigns peaked. Kanabi communities were vilified as “foreigners” and “RSF supporters” despite facing abuses from those very forces. These campaigns culminated in real-world attacks and killings of Kanabi residents in January 2025, after the SAF and allied militias regained control of Gezira state.
This case illustrates how algorithmic amplification of hateful, low-credibility narratives does not remain online—it spills into the real world, enabling ethnic scapegoating, incitement, and violence. Without transparency and accountability from tech companies, algorithmic design continues to exacerbate conflict dynamics and expose marginalized communities to existential threats.
4. Incontextual Policy Design Leading to Platform Migration
Beyond enforcement failures, the design of some content moderation policies creates additional harm. For instance, Meta’s Dangerous Organizations and Individuals (DOI) policy, which bans content from groups on the U.S. Foreign Terrorist Organizations (FTO) list, also restricts political communications and public updates in contexts like Syria. As a result, groups like Hay’at Tahrir al-Sham (HTS) and Syria’s interim leadership have turned to Telegram, where moderation is minimal and misinformation thrives. This migration to less regulated platforms opens up space for disinformation to spread unchecked, worsening the conflict’s information environment.
The same DOI policy was used by Meta’s oversight board to force the company to overturn an automated decision to keep a video that shows armed individuals detaining a person in the back of a military vehicle. One man, speaking in Arabic, identifies himself as a member of the Rapid Support Forces (RSF) and states that they have captured a foreign national, presumably a combatant linked to the Sudanese Armed Forces (SAF). He adds that the detainee will be handed over to RSF leadership and expresses the group’s intent to pursue and apprehend SAF leaders along with any foreign affiliates present in the country. According to Memonic, removing such videos under Meta’s DOI policy risks silencing vital, life-saving information that civilians rely on to track troop movements, locate loved ones, and avoid danger. In Sudan, content posted by actors like the RSF often reveals the treatment and whereabouts of hostages or provides clues about imminent threats. While moderation should prevent harm, blanket takedowns erase evidence, hinder public awareness, and obstruct humanitarian response. Transparency and context-sensitive moderation are urgently needed.
Additionally, some tech companies have gone so far as to permanently suspend fact-checking organizations’ accounts on their platforms. One notable example is the Palestinian Platform for Fact-checking and Media Literacy, known as Kashif—a member of the International Fact-Checking Network (IFCN)—whose account, dedicated to verifying news and information related to the occupied Palestinian territory (oPt), was suspended on TikTok at the beginning of this ongoing genocide. This action has further exacerbated the challenges faced by fact-checking organizations in Palestine, particularly amid ongoing escalations on the ground, where access to accurate, life-saving information is more critical than ever.
Flawed and inconsistent content moderation—whether through over-enforcement, under-enforcement, or poor policy design—has become a key driver of information harm in conflict-affected regions. These failures not only obstruct access to truth and justice but also put real lives in danger by disrupting the flow of critical information when it’s needed most. Addressing these issues requires greater transparency, contextual expertise, and collaboration with local civil society to design moderation frameworks that prioritize human rights and the social and political realities of communities on the ground.
Shifting Tech Company Policies: Highlighting how recent changes by tech companies are affecting the way platforms handle content in sensitive or high-risk contexts
All of the above indicates that tech companies’ existing efforts have already fallen short in enabling access to life-saving information during times of crisis—and, in many cases, have contributed to endangering lives in complex contexts. Yet earlier this year, tech companies such as Meta began implementing major policy changes, including—but not limited to—replacing human experts with AI to assess privacy and societal risks, terminating their fact-checking programs, and weakening their hate speech policies.
These changes are extremely alarming and could further endanger lives. Replacing human specialists with AI to assess societal risks is especially dangerous, as AI lacks the contextual, cultural, and ethical understanding required to evaluate complex, real-world harms. It cannot recognize power imbalances, interpret nuanced threats like election interference or gender-based abuse, or make informed decisions rooted in human rights. AI systems are not neutral—they often amplify existing biases embedded in the datasets they are trained on. As long as companies refuse to allow public audits of these datasets, concerns about bias and accountability will remain unresolved. Without human oversight, automated content moderation decisions become less transparent and more prone to error, ultimately undermining trust and increasing the risk of serious harm—particularly in sensitive or high-risk environments.
A new report launched by UltraViolet, All Out, and GLAAD, highlights the severe consequences of Meta’s rollback of its U.S. fact-checking program and hate speech policies. Drawing on a survey of 7,000 active users—primarily women and LGBTQ+ people—across Facebook, Instagram, and Threads, the findings are alarming: 1 in 6 respondents reported experiencing gender-based or sexual violence on Meta platforms; 92% expressed concern about increased harmful content; 72% observed more hate targeting protected groups; and 77% now feel less safe expressing themselves. Additionally, 66% have witnessed harmful content in their feeds, while over a quarter say they’ve been directly targeted with hate or harassment. These results underscore the tangible harm following Meta CEO Mark Zuckerberg’s sweeping policy changes, announced on January 7, 2025, just prior to Donald Trump’s inauguration.
The recent changes in the tech sector didn’t happen overnight. Tech companies have been steadily laying off employees over the past few years, often justifying these decisions under the guise of increasing efficiency. However, given the growing risks and documented harms caused by their technologies—particularly in the Global Majority and conflict-affected regions—these layoffs represent more than just cost-cutting measures. They signal a troubling deprioritization of people in marginalized contexts and a failure to adequately address the escalating societal and political risks their platforms contribute to.
Furthermore, the rise of digital authoritarianism—a phenomenon long entrenched in many Global Majority countries and now increasingly visible even in democratic contexts like the United States—adds another layer of urgency. The convergence of shrinking human oversight, the growing use of AI and other technologies for surveillance and military purposes, opaque algorithmic decision-making, and weakening policy safeguards demands a fundamental reassessment of tech accountability. We must urgently ask: Who is responsible when harm occurs, and how can meaningful oversight be restored in systems that are becoming increasingly automated, opaque, and unaccountable?
While some countries in the Global Majority have taken the lead in developing new strategies to hold tech companies accountable, others lag behind. For instance, in June, Brazil’s Supreme Federal Court issued a landmark ruling expanding tech platforms’ legal responsibility for user-generated content, requiring them to proactively monitor and remove posts related to “serious crimes.” This decision sets a new global precedent for platform accountability and challenges the notion that tech companies are merely neutral intermediaries.
In contrast, in the SWANA region—where the consequences of tech companies’ failures are particularly severe for those living under oppressive regimes or in conflict-affected areas—content moderation policies are often inconsistently applied. Vague or biased enforcement mechanisms lead to the over-removal of critical content while allowing hate speech, incitement to violence, and disinformation to spread unchecked, particularly against marginalized communities. These issues are further compounded by the continued failure of tech companies to fix or adequately localize their fact-checking systems, leaving harmful content unaddressed while silencing those trying to speak out.
Policy recommendations
This analysis demonstrates that voluntary transparency and self-governance by tech platforms are insufficient. There is an urgent need for mandatory global regulation that ensures no one is left behind. Accordingly, we offer the following recommendations for policymakers:
- Reversing Policy Changes by Tech Companies: The companies must be pressured to reverse recent policy changes related to data access on their platforms, the dismantling of fact-checking programs, the weakening of hate speech policies, and the automation of risk analysis, among other measures.
- Enhancing Transparency in Platform Operations: Policymakers must mandate tech companies to disclose detailed information about their content moderation practices and algorithmic systems, including how content is ranked—especially during crises. Platforms should also be obligated to provide greater transparency around both government and voluntary content removal requests, including how these requests are handled.
- Enabling Independent Audits of Algorithms and Moderation Practices: Companies must allow independent third-party audits of their moderation and recommendation algorithms, with particular attention to high-risk regions. Content removed during conflicts should be thoroughly documented, subject to appeal, and publicly reported in near real-time.
- Conducting Human Rights Impact Assessments: Governments and intergovernmental bodies (such as the UN) should compel tech companies to conduct rigorous, public-facing human rights impact assessments in all regions where their platforms operate—especially in conflict zones. These assessments should evaluate how content moderation policies affect marginalized communities and must also be conducted whenever companies launch new products or implement policy changes.
- Creating an Independent Digital Platforms Accountability Council: Regional and international bodies should explore and pilot the creation of an Independent Digital Platforms Accountability Council, body, or commission—modeled after or expanding on entities like the Digital Services Coordinators in the EU or the Meta Oversight Board—but with full financial and administrative independence from tech companies.
- Implementing Crisis Response Protocols: Mandate that platforms activate crisis response protocols during war, prioritizing safety, factual information, and public access. These protocols should be published publicly to enhance transparency and user notification regarding content removals or visibility reductions. They should also elevate credible, verified local sources and humanitarian guidance, while minimizing the invasive use of automation in content moderation.
- Strengthening Partnerships with Local Fact Checkers: Policymakers must mandate tech companies to support and collaborate with local fact-checkers and journalists to enable fact-checkers and media outlets to operate freely on digital platforms, this includes creating safeguards against arbitrary suspensions of verified fact-checking organizations, providing funding to support independent, locally grounded fact-checking networks in under-resourced regions like SWANA.
- Investing in Contextualized and Localized Moderation Systems: Regulators should mandate that platforms increase their investment in contextualized, localized moderation systems that reflect local political, cultural, and linguistic realities. This includes employing native speakers and regional experts on content review teams, co-designing flexible moderation policies that take political complexity into account and distinguish between documenting atrocities and promoting violence. Platforms should also be required to provide transparency around language coverage, enforcement rates, and the accuracy metrics of AI moderation tools—modeled on the Digital Services Act (DSA) transparency requirements, but applied globally.