Facebook parent Meta’s systems fail to block anti-Muslim propaganda advertisements amid ongoing Lok Sabha elections, claims report

In 2020, Facebook parent Meta faced heavy criticism for allowing the "Stop the Steal" movement to spread like wildfire on its platform. This far-right movement falsely claimed that electoral fraud had deprived Donald Trump of victory in the 2020 United States presidential election, leading to widespread violence in the country. Now, Meta has come under fire again for a similar role in promoting political polarisation during India's ongoing Lok Sabha elections.

A recent research report by the corporate accountability group Ekō, in collaboration with India Civil Watch International, has revealed alarming shortcomings in Meta's ability to detect and block AI-generated ads that promote hate speech, election disinformation, and incitement to violence. These findings are particularly concerning given the critical election period in India, which several experts have termed as the most important Lok Sabha elections in recent times.
Infiltrating election silence periods
The investigation, spanning the third and fourth phases of India's seven-phase election, targeted 189 contentious constituencies during the election's 48-hour "silence period." This period mandates a halt on all election-related advertising, in order to grant citizens some breathing room to make a considered voting decision. However, researchers found that Meta failed to enforce these restrictions, allowing the dissemination of harmful political advertising, mainly by the ruling Bharatiya Janata Party (BJP).
Type of ads approved by Meta
Between May 8-13, the study organisations received publishing approvals for 14 self-submitted highly inflammatory advertisements, meaning that Meta’s auto-check systems were unable to pick up on the posts’ disturbing content. These ads contained violent rhetoric anywhere from targeting Muslim minorities, spreading disinformation by exploiting communal conspiracy theories, and inciting violence through Hindu supremacist narratives. One ad even mimicked a recent AI-doctored video of Home Minister Amit Shah, which had recently gained enough notoriety to sanction arrests of some opposition party members.

The approved ads were placed in English, Hindi, Bengali, Gujarati and Kannada, and mimicked or exaggerated existing Hindu-supremacist propaganda, such as:

• Ads targeting BJP opposition parties with accusations of Muslim favouritism
• Dog whistles about Muslim “invaders” and calls for violence against Muslims
• Narratives claiming electronic voting machine tampering, calling for uprisings
• Hindu supremacist language and calls for burning Muslims
• A call for executing a lawmaker accused of Pakistani allegiance
• Disinformation about affirmative action policies

Each ad featured manipulated images created using common AI tools such as Stable Diffusion, Midjourney, and Dall-e, highlighting how easily AI can amplify harmful content.

This news comes despite Meta’s earlier announcements to detect and label AI-generated content to protect the electoral process. Of the total 22 provocative ads that had been submitted to the platform, 14 were reportedly approved within 24 hours, despite them all breaching Meta’s own policies on hate speech, misinformation, and violence.
Exploiting Meta's weaknesses
The research also highlighted vulnerabilities in Meta’s automated ad review system. When faced with regional roadblocks, researchers managed to simply set up dummy Facebook accounts from outside India to circumvent Meta's security measures and post these ads — including during the silence period. Even in the rare chance that these ads were flagged by the platform, slightly adjusting the content was enough to bypass Meta’s scrutiny, the researchers found. This reveals how easily bad actors can exploit Meta’s platform to spread disinformation and incite violence.

Civil society groups, both in India and internationally, have called on Meta to enforce stricter measures during elections, emphasising the need for transparency and accountability. As India continues its electoral process, the need for robust action from social media giants like Meta is more critical than ever to protect democratic integrity and public safety.

A comprehensive report of the investigation can be accessed here.