SMEX and the civil-society campaign on Meta's Arabic-language content moderation (2020–ongoing)

The campaign anchored by SMEX — the Beirut-headquartered Lebanese digital-rights organisation — and built up with Access Now, 7amleh, Article 19, Human Rights Watch, the Electronic Frontier Foundation, and over 70 further civil-society partners is the corpus's principal grassroots-civil-society response to structural bias in Meta's AI-mediated Arabic-language content moderation. Its documentary case — first assembled in late 2020 and confirmed by an independent human-rights due-diligence audit in September 2022 — is that Meta's automated content-moderation systems on Facebook and Instagram systematically discriminate against Arabic-language users: an Arabic hostile-speech machine-learning classifier with no Hebrew-language equivalent; confidence-threshold reductions to 25% for Palestinian content and 40% for broader MENA content; a Dangerous Organizations and Individuals list composed 70% of Arabic-speaking entities; and documented error rates of up to 77% for the Arabic-language classifier on non-violent content. Distinct from the 7amleh-anchored Palestinian-content-moderation campaign — which is the corpus's wartime response to post-October 2023 over-enforcement of Palestinian content specifically — the SMEX campaign is the WANA-regional structural-advocacy counterpart: longer-running (2020 to present), broader in geographic scope (Lebanon, the wider Arab region, and the cross-WANA advocacy field), and rooted in the operational documentation of individual cases alongside the policy-advocacy strand.

Origins: the December 2020 open letter and the structural bias case

The campaign's documentary opening is SMEX's October 2020 article "How Does Palestinian Content Counter the Bias of Social Media Companies?", which surfaced Sada Social Center research documenting more than 950 violations against Palestinian users between 2019 and 2020, named Facebook's 85% compliance rate with Israeli government deletion requests, and identified AI algorithms automatically flagging Arabic political speech terms — martyr, resistance, Hamas, Islamic Jihad — regardless of analytical or contextual use. The campaign's first coordinated multi-organisation action followed on 17 December 2020, timed to the tenth anniversary of the Arab Spring: an open letter to Facebook, Twitter, and YouTube co-signed by over 50 civil-society organisations laying out a five-point demand set — Arabic-dialect moderation expertise; regional civil-society co-design of content policy; implementation of the Santa Clara Principles on moderation transparency; distinction between abuse documentation and glorification of extremism; and functioning transparent appeals. The letter named as the structural root cause not individual moderation decisions but the platforms' failure to develop adequate Arabic-dialect expertise and to engage MENA and WANA civil society in policy design — a framing that has remained the campaign's analytical anchor through five years of escalation and coalition expansion.

The May 2021 escalation and the Oversight Board

The Sheikh Jarrah evictions and May 2021 Israel-Gaza military operations produced a documented mass-removal wave that gave the structural case an incident backbone. Between 6 and 19 May 2021 alone, Instagram removed or restricted at least 250 pieces of Palestine-related content and Facebook at least 179, and SMEX — in coordination with 7amleh and regional partners — launched the #SavePalestinianVoices petition demanding that Meta immediately stop censoring Palestinian content, cease automated AI-based moderation targeting Palestinian content specifically, and publish all Israeli Cyber Unit takedown-request compliance data. In parallel, SMEX's Digital Safety Helpdesk — launched in April 2020 to support journalists, activists, and human-rights defenders across Arabic-speaking countries — handled more than 100 cases in 2021, with 57 Facebook incidents and 27 Instagram incidents documented and approximately 50 escalations direct to platforms; SMEX characterised Facebook's Arabic-language moderation practice as sloppy moderation, especially in regards to Arabic content. On 24 September 2021, SMEX led a nine-organisation joint statement — including Access Now, Article 19, 7amleh, the Electronic Frontier Foundation, Mnemonic, and others — welcoming the Facebook Oversight Board's ruling that Facebook's removal of an Al-Jazeera Gaza post was an unjustified restriction of freedom of expression, and using the ruling to press five structural demands: an independent audit of content-moderation policies on Palestine; disclosure of government takedown requests and compliance rates; transparency on automation and algorithms in content moderation; publication of internal procedures for dangerous-groups classification; and co-design of improved policies with civil society. The 2021 advocacy round positioned the Oversight Board's formal accountability infrastructure as the primary multistakeholder pressure-point — a posture distinct from the litigation-side approach of the corpus's other content-moderation campaigns.

The BSR Human Rights Impact Assessment and the 71-organisation response (2022)

The campaign's structural case received its most consequential external validation in September 2022. The Business for Social Responsibility (BSR) Human Rights Due Diligence report on Meta's conduct during the May 2021 escalation — commissioned by Meta following a Meta Oversight Board recommendation — confirmed 21 specific reform recommendations and documented, on the record, the core claims the campaign had been making since 2020: Arabic content was flagged at significantly higher rates than Hebrew and was more likely to be erroneously removed; Meta had deployed an Arabic hostile-speech ML classifier with no Hebrew-language equivalent; confidence thresholds had been lowered to 25% specifically for Palestinian content and 40% for other MENA content; and Meta's Dangerous Organizations and Individuals list was 70% composed of Arabic-speaking entities. On 27 September 2022, SMEX and 71 partner organisations — including Human Rights Watch, the Electronic Frontier Foundation, Access Now, Al-Haq, and numerous Palestinian and regional groups — issued a joint statement demanding full and timed BSR-recommendation implementation, complete transparency on Israeli government content-removal requests, disclosure of automated decision-making in content moderation, creation of a Hebrew hate-speech classifier matching the existing Arabic one, and a civil-society co-design process. The September 2022 joint statement marks the campaign's widest coordinated mobilisation and the moment at which the over-enforcement pattern shifted from documented allegation to independently-confirmed finding — with Meta as the commissioning party of the audit that confirmed it.

#StopSilencingPalestine and the MENA Alliance for Digital Rights (2023–ongoing)

The post-October 2023 period saw the campaign scale up into a named coalition infrastructure. SMEX's #StopSilencingPalestine campaign — coordinated with Access Now, Article 19, 7amleh, Human Rights Watch, the Electronic Frontier Foundation, Amnesty International, Witness, and 20+ additional organisations — brought coalition representatives into direct meeting with Meta's President of Global Affairs Nick Clegg and senior trust-and-safety officials in February 2024. On 14 March 2024 the coalition issued a formal letter to Meta; Meta's April 2024 written response was rejected by coalition members as failing to "outline in concrete terms the steps it has taken to identify the negative impacts of its policies." No substantive Meta-policy concession on the DOI policy, the Arabic/Hebrew-classifier asymmetry, or the lowered confidence thresholds has been achieved since the BSR audit. Alongside the coalition campaign, SMEX helped found the MENA Alliance for Digital Rights (MADR) — a coalition that includes 7amleh, Mnemonic, Access Now, the Electronic Frontier Foundation, and Article 19 and that has issued open letters to Meta and TikTok while maintaining a public content-moderation policy FAQ documenting the campaign's structural technical claims: the 77% Arabic-classifier error rate on non-violent content, the Hebrew hate-speech classifier gap, the DOI list composition, and — surfaced at Bread&Net 2021 — the figure that Meta served 220 million Arabic speakers across 30+ dialects with only 766 Arabic-content moderators. MADR is the campaign's durable coalition infrastructure for the post-BSR enforcement phase.

Lebanon and feminist digital rights: the non-Palestinian dimension (2025)

SMEX's April 2025 research report "Confronting Structural Silencing: Challenges and Resistance Among Digital Feminist Activists in Lebanon" confirmed that the structural Arabic-language content-moderation bias the campaign has been documenting since 2020 extends beyond Palestinian political speech to Lebanese feminist civil-society work. The research — conducted with nine Lebanese feminist activists — documented Meta's over-removal of Arabic-language feminist content while misogynist content targeting women remained visible, identified algorithmic awareness as a new dimension of digital literacy activists must navigate, and grounded the structural case in Lebanese civil society rather than in the wartime Palestinian context that has dominated the campaign's most visible documentation. This dimension is what distinguishes the SMEX campaign from the 7amleh-anchored Palestinian-content-moderation campaign: where 7amleh documents asymmetric wartime over-enforcement of Palestinian content specifically, the SMEX campaign's structural register is the cross-WANA, cross-issue pattern — that Meta's Arabic-language AI-mediated moderation systems are miscalibrated against Arabic-language civil-society and political speech as such, regardless of whether the subject is Palestinian rights, Lebanese feminist advocacy, or Arabic-language sexual and reproductive health rights content.

Significance for the broader AI-good movement

The SMEX campaign is the corpus's principal grassroots-civil-society engagement with the Arabic-language dimension of AI-mediated content moderation. Structurally, it occupies the WANA-regional position inside a global field where the corpus's other platform-accountability campaigns are anchored in the UK (the stop-live-facial-recognition coalition), the United States (the Athena Coalition, the Concept Art Association, the SAG-AFTRA AI provisions), Latin America (Coding Rights on facial recognition in Brazil, Derechos Digitales' IACHR hearing), and East Africa (Foxglove and the Kenyan content-moderator cases). The SMEX campaign brings that field its Arabic-language, WANA-anchored counterpart — with the structurally distinctive feature that the AI system being contested is a classifier the platform built targeting Arabic-language speech, not a general-purpose system that produces Arabic-language disparate impact incidentally. The classifier asymmetry — an Arabic hostile-speech ML system with no Hebrew-language equivalent, developed by a US-headquartered platform serving a predominantly non-US Arabic-speaking user base — is the campaign's clearest demonstration of how AI content-moderation infrastructure encodes the priorities and blindspots of the context in which it was built. The BSR audit's confirmation that this is a design-and-resourcing pattern rather than an inadvertent technical failure, and the coalition's sustained failure (2022–2025) to translate that confirmation into a substantive Meta-policy concession, frames the campaign as the corpus's standing case of a structural AI-good diagnosis documented and confirmed but not yet remediated.