Nearly a decade after social media campaigns fuelled atrocities against the Rohingya minority, a new investigation reveals that the Meta Facebook platform’s content moderation systems are failing to protect vulnerable groups in Myanmar.
The exhaustive study, conducted over a year by the monitoring group Human Rights Myanmar (HRM), found that despite reporting clear instances of actionable hate speech, over a third of the flagged content remained publicly accessible after a month.
The research, supported by the Pulitzer Center, highlights a systemic gap between Meta’s public safety commitments and the reality on the ground in conflict-affected environments.
Rohingya remain the primary target
According to the report, “actionable” hate speech—expression that may or must be restricted under international law—remains persistent in Myanmar’s online space. The Rohingya continue to bear the brunt of this abuse, targeting them in 57 per cent of all actionable hate speech identified on Facebook, while Muslims generally were targeted in 33 per cent of the analyzed posts.
Crucially, the study notes that modern online hostility rarely takes the form of explicit calls to violence, which accounted for just one per cent of the monitored posts on Facebook.
Instead, it is dominated by targeted identity mockery and slurs, at 75 per cent, and collective guilt narratives at 61 per cent. These narratives systematically criminalize and dehumanize entire civilian populations, contributing to a broader environment of fear and hostility.
The failure of user reporting and suspected ‘auto-Rrejections‘
The investigation tested Facebook’s systems by observing 232 actionable hate speech posts over a month. While Facebook’s proactive automated moderation successfully deleted 57 per cent (132 posts) of the content before any intervention, the platform’s manual reporting tools proved almost entirely ineffective.
When HRM reported the remaining 100 publicly accessible posts using standard user accounts, Facebook accepted only three per cent of the reports, rejecting 89 per cent and leaving 8 per cent pending indefinitely. Subsequent appeals of the rejections yielded similarly meager results, with 76 per cent rejected and nine per cent left unanswered.
In total, 36 per cent of the original hate speech remained online after a month, the HRM investigation finds.
Even more troubling, the rejections followed an incredibly rigid timeline, raising suspicions of automated backlog-driven rejections. A striking majority of the reports were rejected after exactly 96 or 120 hours (four and five days), while appeals were rejected precisely 156 hours (6.5 days) after submission.
“The repeated timing pattern raises legitimate questions about whether some reports were rejected through administrative workflow rather than substantive assessment,” the report states, suggesting that Meta’s systems may automatically reject backlogged complaints as they hit operational deadlines to manage queue sizes.
The ‘More Speech’ policy and the global pullback
The failures observed in Myanmar coincide with a massive global reduction in Meta’s content moderation efforts. In January 2025, CEO Mark Zuckerberg announced a new “more speech, fewer mistakes” approach, scaling back proactive automated moderation in favor of avoiding mistaken removals.
The real-world impact of this policy shift is laid bare in Meta’s own transparency data:
- Violence and incitement deletions plummeted globally from 11.5 million in the second half of 2024 to just 1.4 million in the second half of 2025.
- Hateful conduct deletions dropped from 12.2 million to 2.5 million over the same period.
A call for heightened due diligence
Under the U.N. Guiding Principles on Business and Human Rights (UNGPs), social media platforms are required to conduct context-specific human rights due diligence, particularly in high-risk, conflict-affected areas like Myanmar.
HRM concludes that Meta has not yet demonstrated that its systems effectively mitigate these foreseeable harms. The group called on social media companies to publish country-specific enforcement data, implement heightened due diligence, and conduct rigorous human rights impact assessments before implementing sweeping global policy changes.
“Myanmar has repeatedly demonstrated the consequences of allowing online hatred to circulate unchecked,” the report warns, urging immediate structural reforms before online rhetoric translates once again into offline tragedy.
A double-edged sword in Myanmar’s information ecosystem
In Myanmar’s volatile human rights environment, Facebook occupies a unique and incredibly powerful role, acting as a double-edged sword that can either mitigate or severely exacerbate societal crises.
While the platform has the potential to facilitate free public debate and promote civic accountability, it has also repeatedly served as a major vector for community polarization, spreading systemic distrust, intolerance, and hatred.
Crucially, the U.N. Independent Investigative Mechanism for Myanmar (IIMM) concluded that coordinated online hate campaigns on Facebook played an important role in spreading hatred, dehumanizing the Rohingya, and cultivating hostility preceding and accompanying the military’s 2017 atrocities.
Following intense international criticism for its role in those clearance operations, Facebook’s parent company acknowledged it had not done enough to prevent its platform from contributing to offline violence and committed to strengthening its human rights due diligence and moderation safeguards in the country—promises that researchers warn are now under threat.
