Deepfake and cheapfake fears in elections, 10 tactics used more

It was certain which party would win the most seats in Bangladesh’s 12th Jatiya Sangsad election held on 7 January 2024. The election became known as a ‘dummy vote’.

On the morning of the polling day, however, a fake or ‘deepfake’ video of Abdullah Nahid Nigar, an independent candidate in the Gaibandha-1 constituency, went viral. In the video, he was seen saying that he has withdrawn from the election, which confused many voters.

This is one Bangladeshi example of the use of deepfake videos in elections. A 2024 report titled The Influence of Deepfakes on Elections by the German organisation Konrad Adenauer Foundation (KAS) says such incidents have occurred frequently in recent elections around the world.

The United States, Turkey, Slovakia, Argentina, Indonesia, India, Poland, Bulgaria, Taiwan, Zambia and France the list of such examples goes on.

The KAS report mentions the Bangladeshi case. Another example comes from the United States. In January last year, ahead of the Democratic Party primaries in the state of New Hampshire, a video of then president Joe Biden circulated in which he appeared to discourage voters from voting. The video was a deepfake.

The January 2024 incident in Bangladesh was fact-checked by Minhaz Aman, a fact-checker who is now head of research at the organisation Activate Rights.

He told Prothom Alo Saturday, that the misuse of artificial intelligence (AI) in elections has been seen in various countries. Deepfakes can influence voters’ choices in elections and also harm candidates in different ways.

The fact-checker said digital literacy in Bangladesh is in a dire state. If AI-generated videos calling for election boycotts or damaging a candidate’s image are spread here, the impact could be devastating.

At the start of 2024, AI was not as advanced or accessible as it is now. Today, fake videos, audio and images can be created using AI in such a realistic way that they appear completely genuine. This is known as a deepfake.

Another type is ‘cheapfake’, which is created using inexpensive software rather than AI. Cheapfakes are usually used to misrepresent events, such as spreading fake statements by copying photo cards posted on media outlets’ Facebook pages.

From what has been learned from speaking to fact-checkers, analysts and law enforcement officials, 10 tactics are most commonly used to spread disinformation in the country through deepfakes and cheapfakes.

These include attaching misleading or differently interpreted captions to genuine photos or videos, cutting parts of real statements or changing their context to create a different meaning, attributing completely fabricated statements or quotations to specific individuals, and presenting old photos, videos or news as if they were recent events.

As the election draws closer, the scale of political disinformation is increasing. At least 13 top and high-profile leaders from several political parties have already fallen victim to such incidents.

An analysis of misinformation trends from July to September last year was published on 29 October by the fact-checking organisation Dismislab. Based on data from eight local fact-checking organisations, the report said that video was the main medium for spreading false information during that period, accounting for 66 per cent of all fact-checked content.

Compared to the previous three months (April-June 2025), the use of graphics, images and written posts declined. In other words, as the election approaches, the spread of confusion through AI-generated fake videos is increasing.

There have also been incidents of creating fake graphics modelled on media photo cards, television scrawls or news portal designs, using fabricated numbers or statistics, forging screenshots or documents, and coordinated campaigns in which the same false information is spread simultaneously from multiple pages and accounts.

The widespread circulation of various types of misinformation and video-based disinformation in the months leading up to the election has also been highlighted in an analysis by the fact-checking organisation Rumor Scanner.

According to the organisation, 1,441 instances of misinformation were identified between October and December last year. Of these, 956 were political disinformation. Video-based misinformation was the most prevalent, with 651 cases, followed by text-based misinformation at 550 and image-based misinformation at 245.

Rumor Scanner says that by nature, 1,051 of these were outright falsehoods, 253 were distorted information, 132 were misleading, three were partially false and two were partially true. This shows that unchecked and deliberately fabricated falsehoods were the main tools for spreading disinformation during this period.

A Dismislab report published on 6 January says that between 16 November and 15 December, nine fact-checking organisations in Bangladesh identified and debunked 63 false election-related claims. In the previous month, from 16 October to 15 November, the number was 50, meaning false claims increased by around 26 per cent. During this time, fabricated statements and quotations were used the most.

Analysts say that some individuals based abroad regularly spread falsehoods and disinformation about Bangladesh on Facebook, YouTube and other platforms, mixing truth and lies and inciting unrest. This is a widely known fact. However, it is often not possible to identify who are producing coordinated fake videos, images and audio.

Miraj Ahmed Chowdhury, managing director of the research organisation Digitally Right, which studies the impact of technology on information systems, told Prothom Alo that identifying the accounts and pages spreading disinformation is not difficult. But determining who is behind them requires investigations that fact-checkers cannot carry out.

He said that, broadly speaking, two types of people are involved in spreading disinformation: one group is politically or ideologically motivated, while the other does it for money. Based on observable patterns, it is not unreasonable to suspect the involvement of political forces behind them.

Fact-checkers analysing trends say there are multiple armies of Facebook pages and accounts opened under fake identities, known as “bot armies”. These groups work on behalf of political parties. Some have stronger networks, while others have less influence.

Law enforcement sources say that disinformation is being spread both for and against parties participating in the election. At the same time, leaders and activists of the Awami League, whose activities have been banned, are creating significant risks by spreading disinformation online despite not contesting the election. Announcements aimed at obstructing the election have also come from various levels of the party.

Speaking on condition of anonymity, an officer of the Detective Branch (DB) of Dhaka Metropolitan Police told Prothom Alo that Awami League leaders and activists are forming groups on Telegram and other online platforms to spread various types of disinformation. Law enforcement agencies have identified hundreds of such pages and groups.

Field-level government officials are also concerned about disinformation. At a meeting of the Election Commission with returning officers and field administration officials on 23 December, challenges related to the misuse of social media and technology and the spread of rumours were raised.

The National Cyber Security Agency (NCSA) has formed a special cell to prevent fake information, misleading content and rumours until the next national election. The NCSA is also coordinating with the Chief Adviser’s Press Wing, the Press Institute of Bangladesh (PIB), Bangladesh Sangbad Sangstha (BSS), the Bangladesh Telecommunication Regulatory Commission (BTRC) and the law enforcement agencies.

The Criminal Investigation Department (CID) of the police is also working to prevent fake information and rumours during the election. Deputy Inspector General (DIG) MA Bashar Talukder of the CID’s Cyber Police Centre told Prothom Alo that social media is being monitored round the clock ahead of the election. They are prepared to counter any disinformation through fact-checking and forensic analysis.

The problem is that disinformation spreads rapidly, while removal or action often takes time. Before that happens, the content reaches people and many come to believe it.

The situation has reached a point where even senior leaders of political parties are making statements based on rumours circulating on social media. At least seven political leaders and prominent figures have spoken on various talk shows citing fake Facebook photo cards as sources. These incidents were later proven to be false.

Some media outlets are also publishing fake news circulating on social media without citing specific sources. For example, two outlets published images claiming to be from the date of BNP chairperson Tarique Rahman’s return to the country. Both images were AI-generated.

Professor BM Mainul Hossain, director of the Institute of Information Technology at the University of Dhaka, told Prothom Alo, “There are examples around the world where elections have been postponed due to cyber-attacks and deepfakes. In our country too, the Election Commission has repeatedly warned about the risks of deepfakes and disinformation. But effective counter-measures are not very visible.”