Facebook AI failed to block Brazilian electoral disinformation ads
Amnesty International Brasil tested Facebook's ad moderation system ahead of Brazil's 2026 elections and found it failed to block ads containing electoral disinformation, including false claims promoting military intervention. The test submitted 15 advertisements with 14 containing misinformation; Meta's filters approved eight ads with false electoral information and retained seven only due to identity verification gaps rather than misleading content.
What happened
Amnesty International Brasil conducted a test of Meta's Facebook ad moderation system before Brazil's 2026 presidential elections. The organization submitted 15 paid advertisements in Portuguese targeted at Brazilian voters aged 16 and over, with 14 containing electoral disinformation and one neutral.
Meta approved eight ads containing false information about the electoral process from accounts submitting without required identity verification. Seven additional ads were retained only for lack of advertiser identity verification, not for the misleading content itself. Among ads that passed, three of four promoting military intervention discourse were accepted, including one stating: 'We need a military revolution, do not vote in the elections, and the military will retake power if less than 50% vote.'
Reported impact
The test exposed that Facebook's moderation filters showed inconsistent performance, demonstrating greater capacity to block ads accusing candidates of crimes or fraud while allowing approval of content promoting military rhetoric and messages against democratic institutions.
Index analysis
The disinformation strategies in the test ads followed three approaches: delegitimizing elections through undermining confidence in voting systems or results; constructing enemies by portraying political opponents and courts as national threats; and promoting authoritarianism by presenting military intervention as necessary responses to crime or political disorder.
Governance controls
Meta responded that Amnesty International's report considered 'a very small sample of ads' and stated it invests 'significant resources' to protect Brazilian elections. The organization ensured submitted ads were scheduled for future dates and cancelled before publication to prevent actual posting of the content on the platform.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.