AI Misinformation Threatens Integrity of Bangladesh Elections, Watchdogs Warn

In a concerning development ahead of Bangladesh’s February 12 parliamentary election, artificial intelligence-generated campaign videos are flooding social media platforms, raising significant alarms about voter manipulation and electoral integrity.

An investigation by Dismislab, an independent fact-checking organization, has uncovered hundreds of AI-created videos circulating online that show supposed ordinary citizens endorsing political parties. Most troublingly, these videos lack clear disclosure of their artificial nature, potentially misleading voters in one of South Asia’s most populous democracies.

The sophisticated AI-generated content typically portrays elderly citizens, rural women, and working-class individuals appearing to speak directly to viewers while praising specific political parties or electoral symbols. Many of these widely shared clips feature individuals seemingly promising economic security, welfare benefits, or justice under particular political banners—narratives carefully crafted to appeal to undecided or economically vulnerable voters.

Technical analysis of the videos revealed telltale signs of AI generation, including repetitive scene resets and unnatural camera angle changes occurring at regular intervals. Despite these indicators, the videos are being presented as authentic testimonials from real voters and distributed as legitimate campaign material without proper labeling.

Between January 1 and 15 alone, Dismislab documented more than 800 AI-generated political videos across Facebook, YouTube, and TikTok. On Facebook, researchers traced 576 such videos to 21 separate pages, with approximately 60 percent carrying no indication whatsoever that they were produced using artificial intelligence.

Even in cases where disclosure exists, it has been inconsistently implemented. The investigation found that Facebook’s AI information label appeared on some versions of its mobile app but disappeared when the same videos were viewed through Facebook Lite or desktop browsers. Similar labeling gaps were identified on other platforms, where the vast majority of analyzed videos lacked any AI-related warnings.

The content promotes various political actors using fabricated characters—including police officers, journalists, religious minority community members, and even children—to enhance perceived credibility and emotional appeal. This strategy appears designed to create the impression of broad-based support across diverse segments of Bangladesh’s society.

Bangladesh’s Election Commission has publicly expressed concerns about AI-driven misinformation multiple times in recent weeks, warning that these emerging technologies could potentially be misused to manipulate voters ahead of polling day. With over 169 million people and relatively recent adoption of widespread internet access, Bangladesh faces particular challenges in combating digital misinformation.

Researchers involved in the investigation have emphasized that low digital literacy levels throughout the country significantly increase the risk that voters may mistake AI-generated content for genuine testimonials. They also noted that during heated election periods, confirmation bias plays a powerful role, with users more likely to trust and share content that aligns with their existing political preferences while ignoring verification steps.

The proliferation of such videos in Bangladesh comes as political tensions remain high in the country. Prime Minister Sheikh Hasina’s Awami League has governed for the past 15 years, and opposition groups have accused the government of increasingly authoritarian tactics. In this charged environment, misleading campaign materials could further inflame divisions or undermine trust in the electoral process.

As generative AI tools become cheaper, more accessible and increasingly realistic, analysts note that Bangladesh’s election highlights a broader global challenge: how democratic systems worldwide can respond to rapidly evolving technologies that blur the line between authentic political discourse and synthetic persuasion designed to manipulate public opinion.

The situation in Bangladesh may serve as a warning for other democracies facing elections in 2024, including India, South Africa, and the United States, where similar AI-generated content could potentially influence voter behavior on a much larger scale.