Facebook allegedly did not adequately remove hate speech, some of which was extremely violent and dehumanizing, on its platform including through automated means, contributing to the violence faced by ethnic communities in Ethiopia.
.avif){kind=small}
Facebook's recommendation algorithms and content moderation failures in Ethiopia have been linked to the amplification of hate speech, disinformation, and incitement to violence against ethnic groups. Internal documents and legal filings allege that the platform's engagement-based ranking systems prioritized inflammatory content, while the company failed to provide adequate moderation resources for local languages like Amharic, Oromo, and Tigrinya. This environment contributed to real-world atrocities, including the targeted murder of individuals and widespread ethnic violence.
Facebook allegedly did not adequately remove hate speech, some of which was extremely violent and dehumanizing, on its platform including through automated means, contributing to the violence faced by ethnic communities in Ethiopia.
The primary risk and reported harm stem from Facebook's algorithms exposing users to and actively amplifying extreme hate speech, genocidal rhetoric, and direct incitement to violence against ethnic minorities.
The real-world harms were driven by the post-deployment operations of Facebook's engagement-ranking and recommendation algorithms, which unintentionally amplified toxic content as a systemic byproduct of maximizing user engagement.
Unacceptable Risk: The reports describe an AI system (engagement-based ranking and recommendation algorithms) used for behavioral influence and manipulation that prioritized and amplified hate speech and incitement to violence, leading to severe physical harm and systemic violations of fundamental rights.
Highest direct severity in any category: Catastrophic. Severity is scored from Negligible to Catastrophic in each harm category, for harm the reports describe as caused directly or indirectly by the AI system.
Reported: The reports explicitly describe numerous fatalities and physical violence resulting from unmoderated incitement on the platform.
Directly caused: N/A
Indirectly caused: Professor Meareg Amare Abrha was shot dead outside his home after being targeted by viral Facebook posts. Over 160 people were killed in mob violence following the death of Hachalu Hundessa, which was actively fueled by Facebook posts. A Qemant village was pillaged and burned down following a false post on Facebook.
Inferred additional harm: It is highly likely that thousands of additional deaths and injuries in the Ethiopian civil war were indirectly fueled or accelerated by the unchecked spread of genocidal rhetoric and ethnic incitement on Facebook.
Reported: The reports explicitly describe the destruction of community infrastructure.
Directly caused: N/A
Indirectly caused: An entire Qemant village was pillaged and burnt down, destroying local homes and community infrastructure, following a false Facebook post alleging terrorism.
Inferred additional harm: Widespread destruction of public and private infrastructure occurred during the ethnic clashes and military offensives that were coordinated and incited via the platform.
Reported: The reports explicitly describe property damage linked to online incitement.
Directly caused: N/A
Indirectly caused: The burning and pillaging of a Qemant village resulted in the total loss of homes and personal property for thousands of fleeing civilians.
Inferred additional harm: Extensive property damage, including looted businesses and destroyed homes, occurred during the broader civil unrest fueled by unmoderated platform content.
Reported: The reports describe a massive legal claim seeking financial restitution for victims.
Directly caused: N/A
Indirectly caused: The lawsuit filed in Kenya's High Court seeks the establishment of a 200 billion Ksh (approximately $1.6 billion USD) restitution fund for victims of hate and violence incited on Facebook.
Inferred additional harm: Displaced families and destroyed businesses suffered catastrophic, unquantified financial ruin due to the violence fueled by the platform's algorithmic amplification.
Reported: The reports explicitly describe a massive proliferation of toxic and malicious content.
Directly caused: Facebook's recommendation algorithms actively promoted and amplified hateful, violent, and genocidal posts because they generated high user engagement.
Indirectly caused: Users, diaspora groups, and state-affiliated actors created and shared massive volumes of hate speech, ethnic slurs, and calls for genocide (e.g., calling Tigrayans 'weeds' and 'cancer').
Inferred additional harm: Millions of users across East Africa were exposed to highly toxic, dehumanizing, and violent content due to the platform's systemic moderation failures.
Reported: The reports explicitly describe systemic differential treatment in safety protections.
Directly caused: Meta's safety and moderation systems performed significantly worse for Ethiopian languages due to a lack of classifiers and human reviewers, representing unequal protection compared to Western users.
Indirectly caused: Ethnic minorities faced targeted harassment, doxing, and physical violence due to the platform's failure to enforce safety standards equitably.
Inferred additional harm: Systemic underinvestment in safety resources likely resulted in similar differential treatment and heightened risk for other non-Western, at-risk nations.
Reported: The reports explicitly describe severe violations of fundamental human and civil rights.
Directly caused: N/A
Indirectly caused: The platform's failure to curb incitement directly contributed to violations of the right to life, physical security, and freedom from discrimination, as highlighted in the Kenyan High Court lawsuit.
Inferred additional harm: The systemic amplification of genocidal rhetoric facilitated widespread crimes against humanity and ethnic cleansing, representing a catastrophic failure of human rights protections.
Reported: The reports explicitly describe the erosion of democratic norms and political stability.
Directly caused: N/A
Indirectly caused: State-affiliated actors (such as INSA) used coordinated networks of fake accounts to spread disinformation, polarize the public, and undermine democratic transitions and peace processes.
Inferred additional harm: The polarization of the information ecosystem has caused long-term damage to social cohesion, trust in public institutions, and democratic governance in Ethiopia.
Reported: The reports explicitly describe targeted privacy violations.
Directly caused: N/A
Indirectly caused: Facebook posts doxed Professor Meareg Amare, revealing identifying personal details and his location shortly before he was murdered.
Inferred additional harm: Numerous other activists, journalists, and academics likely had their private information exposed, placing their lives and families at risk.
Reported: The reports explicitly describe severe psychological trauma and distress experienced by victims' families.
Directly caused: Moti Dereje was severely traumatized and depressed after discovering his father's dead body via a viral Facebook post, which remained online for over four years despite repeated reports.
Indirectly caused: Widespread fear, anxiety, and acute distress among ethnic Tigrayans, Amharas, and Oromos who were subjected to dehumanizing campaigns and violent threats online.
Inferred additional harm: Millions of Ethiopian users exposed to graphic images of violence, such as the video of a Tigrayan man burned alive, likely suffered severe, unquantified psychological trauma.
Reported: The reports explicitly describe severe epistemic harm and the spread of misinformation.
Directly caused: Recommendation algorithms created powerful echo chambers, filter bubbles, and information cascades that hardened ethnic divisions.
Indirectly caused: The platform was flooded with unverified rumors, conspiracy theories, and fabricated claims (e.g., false allegations of a bus hijacking or political plots) that eroded shared reality.
Inferred additional harm: The widespread pollution of the information ecosystem has made it extremely difficult for citizens to access credible news, severely hindering peace and reconciliation efforts.
Meta's autonomous recommendation algorithms systematically amplified ethnic hate speech, disinformation, and incitement to violence in Ethiopia. This algorithmic failure severely exacerbated ethnic tensions, directly contributing to mob violence, targeted killings, and widespread civil unrest during a period of civil war, highlighting the profound threat of unmoderated AI systems to societal stability.