Unlabelled AI videos fuel confusion ahead of polls

As the election approaches, social media platforms are flooded with campaign content promoting political parties. Many of the videos circulating on Facebook, YouTube, and TikTok are created using artificial intelligence (AI).

However, in most cases, these videos are not identified as AI-generated, increasing the risk of voter confusion.

Between 1 and 15 January—the month preceding the vote—fact-checking organisation Dismislab identified around 800 such AI-generated videos. The organisation published an analysis of these videos on its website yesterday, Saturday. It found that 60 per cent of the AI videos posted on Facebook carried no label or identifier indicating that they were AI-generated.

According to Dismislab’s analysis, 576 AI-generated videos were posted from 21 Facebook pages over the 15-day period. Similar content was also posted from seven YouTube channels and two TikTok accounts.

Of the 181 videos analysed from seven YouTube channels, 94.48 per cent had no AI label. None of the 50 videos posted on TikTok carried such a label.

Dismislab also identified a serious inconsistency on Facebook: while an “AI info” label appears when these videos are viewed on the main Facebook app, the label does not appear when the same videos are viewed on Facebook Lite or on computers.

Risk of Influencing Votes Through Confused Voters

Global research and Dismislab’s findings suggest that AI-generated videos can influence voter opinions. Citing comment sections of two AI-generated political videos, Dismislab reported that many commenters believed the videos to be footage of real events.

In one example, an AI-generated video promoting an Islamist political party shows a woman describing the party’s electoral symbol as a symbol of justice. In the video, she is seen praising the party as one that will ensure justice and equality for all.

In the comments section, one viewer wrote, addressing the woman, “Masha’Allah, little sister. May Allah grant you understanding of the deen. I pray for you.”

Under another AI-generated video supporting a different political party, a commenter wrote, “Everyone come forward to make the national leader the next prime minister.”

In addition to attacking political opponents, these videos are also spreading false information.

Suman Rahman, Dean of the School of Social Sciences at the University of Liberal Arts Bangladesh and Head of the Media Studies and Journalism Department, said such videos would undoubtedly influence voters.

“In elections, we tend to take sides,” he said. “If AI-generated content aligns with my political position, I will naturally believe it and share it with others. Even when I see something negative about my opponent, I often believe it without verification—because that is what I want to believe: that they are bad.”

Violation of Law and Guidelines

The Representation of the People Order states that in this election, using artificial intelligence to create, publish, promote, or share false, misleading, biased, hateful, obscene, indecent, or defamatory content intended to mislead voters is prohibited. The election code of conduct also bans the use of AI to spread misleading or false information.

Quoting Election Commission Public Relations Officer Md Ruhul Amin Mallik, Dismislab said that spreading false information would be considered a dishonest intent.

Lawyer Jyotirmoy Barua told Dismislab that the Election Commission should issue clear directives mandating the labelling of AI-generated content.