With only hours left before the country votes in its 13th parliamentary election, concerns are growing over the unchecked spread of artificial intelligence-generated misinformation on social media, as the government and Election Commission struggle to respond effectively.
Experts warn that AI-generated images, videos and audio clips could trigger serious disruptions at any point before vote counting or the announcement of results. They say the technology is increasingly being used as a tool for character assassination against rival candidates, deepening a crisis of trust among voters.
Recent incidents have underlined the scale of the threat. A video circulated online in the name of Zubaida Rahman, the wife of BNP Chairman Tarique Rahman, appeared to show her asking for financial assistance via a bKash number.
The clip was shared from a fake Facebook account opened in the name of their daughter, Zaima Rahman, and was viewed by nearly two million users on Facebook alone. Fact-checkers later confirmed the video was AI-generated.
In another case, an image purporting to show a procession led by Bangladesh Khelafat Majlis Ameer Mamunul Haque went viral and was shared at least 5,000 times on Facebook. Subsequent verification found the image had also been created using artificial intelligence.
Manipulated political messaging has extended beyond visuals. One video circulating under Tarique Rahman’s name was found to have distorted his speech by inserting edited remarks originally made by a leader of another party.
More recently, an AI-generated audio clip imitating the voice of a senior political leader spread widely on social media, in which the voice appeared to call for an election boycott. Analysis later confirmed the audio was fake.
Data from fact-checking organisations shows how rapidly misinformation has escalated. Rumour Scanner reported identifying 577 false claims in January alone — the highest monthly figure in its history — with around 81% related to politics. The organisation published 452 fact-check reports during that month.
Its monitoring and evaluation officer, Kawsar Ahmad, said that in the first eight days of February, the group had already identified 171 instances of misinformation, including 113 linked to political issues. He said election-related false information had grown “uncontrollably” in recent months and was intensifying further this month, making the current election more challenging for fact-checkers than previous polls.
Political parties have also traded accusations. BNP election coordination committee spokesperson and party chairperson’s adviser Mahdi Amin said that allegations of long-standing character assassination campaigns against the party by rival political groups persist.
Jamaat-e-Islami assistant secretary general Ahsanul Mahboob Zubair told TIMES of Bangladesh that while misleading campaigns may have short-term effects, the truth ultimately becomes clear to the public.
He added that not only EC, but political parties themselves, should have acted more responsibly and cautiously in addressing AI-driven disinformation.
The issue has legal implications as well. Under Article 16(Cha) of the election code of conduct, misleading voters, character assassination, and the dissemination of false or deceptive content are prohibited — regardless of whether such material is created using AI or other means. While AI use in campaigning is not banned outright, aggressive or deceptive content is considered a direct violation of the code.
Digital rights and technology experts argue that the core problem lies in a lack of transparency. Miraj Ahmed Chowdhury, managing director of Digitally Right, told TIMES that while AI can be used for legitimate campaign communication, voter suppression, intimidation, or impersonation pose serious risks. He said all AI-generated content should be clearly labelled so users can easily identify it as such.
However, he noted that enforcing effective labelling largely falls on social media platforms, as EC cannot directly intervene in every piece of online content.
EC says it is operating within limited capacity. Brigadier General Mohammad Azizur Rahman Siddiqui, chief coordinator of the law-and-order coordination cell, said an AI-based scanning platform had been developed with support from UNDP.
Given resource constraints, he said the commission is selectively reviewing only content deemed serious or risky to law and order or political stability.
As polling approaches, experts warn that the surge of AI-driven misinformation and disinformation is eroding trust in the electoral information environment. They argue that without greater transparency, clear labelling and proactive engagement from digital platforms, containing the crisis will remain increasingly difficult.
