| বিশ্লেষণ
AI-Generated Online Content
Targeting Political Parties, Government, and Security Forces
২৮ অক্টোবর ২০২৫
Artificial intelligence (AI) has emerged as a significant tool in the domain of information warfare, notably intensifying the capabilities of propagandists. The accelerated generation and propagation of content pertaining to politically controversial and high-stakes social topics is demonstrably affecting public discourse.
The strategic deployment of propaganda is particularly potent in creating societal polarization and shaping public opinion, especially during critical windows such as pre-election cycles, periods of civil unrest, and emergency periods.
AI technologies are being intentionally leveraged for targeted harm, including systematic campaigns of defamation and digital abuse. A related and growing concern is the systematic utilization of AI to proliferate gender-insensitive and misogynistic content across various digital and social media ecosystems.
The Fact-Check and Media Research Team, BanglaFact, of the Press Institute of Bangladesh (PIB) conducted a study of AI-generated online content concerning Bangladeshi issues circulated across social media platforms between July and September 2025. The objective of the study was to identify the dominant themes and political biases embedded in AI-generated videos, with a view to assessing the scope, influence, and potential risks of AI-driven propaganda in Bangladesh.
The report revealed that the majority of AI-generated content—particularly forms of mis(dis)information—targeted political parties, the government, and security forces (police and army). Among the political parties, content related to the Awami League appeared most frequently; however, these materials primarily functioned as positive propaganda in favor of the party, whereas content concerning the government and other political parties was predominantly negative in tone and framing.
Research Methodology
A total of 71 AI-generated content pieces were selected as samples for this analysis. The samples were chosen using a purposive sampling method, focusing on videos considered likely to exert political or social influence across social media platforms. The selection process primarily involved reviewing fact-check reports published by various Bangladeshi fact-checking organizations during the study period (July–September 2025). Additionally, AI-generated videos disseminated through different media outlets within this timeframe were identified using relevant keyword searches.
Content was collected from Facebook, YouTube, Instagram, and X (formerly Twitter). In many cases, the same videos were found circulating across multiple platforms, and several had been fact-checked by more than one organization. The verification of samples utilized a range of AI-detection and media authentication tools, including but not limited to Deepfake-O-Meter, Hive Moderation, Sightengine, Cantilux, and Resemble AI Detect.
Analysis and Description
A total of 71 AI-generated content pieces were analyzed for this study, with each video or image treated as a distinct unit of analysis. The samples were categorized into the following thematic groups: Awami League, BNP, Jamaat-e-Islami, NCP, Interim Government, Army/Police, July Mass Uprising, Accidents, Religious, Gender-Insensitive, DUSCU (Dhaka University Central Students’ Union), and Others. These categories were not mutually exclusive, as a single video could be classified under multiple themes. All percentages were calculated based on the total number of analyzed images/videos (N = 71).
Platform and Medium
The study found that most AI-generated fake news circulated across Facebook, X (formerly Twitter), TikTok, YouTube, Instagram, and Threads. Several content pieces appeared simultaneously on multiple platforms, highlighting the coordinated and multi-platform nature of the disinformation campaigns.
Among these platforms, Facebook exhibited the highest prevalence, with 61 content pieces (86%), followed by TikTok, where 40 content pieces (approximately 56%) were identified. In contrast, X, Threads, Instagram, and YouTube showed a comparatively lower presence of such AI-generated materials.
Table 1- The presence of AI-generated misinformation on social media platforms.
Content Formats
According to the content analysis, video emerged as the most common format for employing AI technology to disseminate propaganda or misinformation. Of the 71 analyzed content pieces, 57 were video-based, while 14 consisted primarily of images.
Overview of Content Categories
The examination of the 71 AI-generated content pieces included in this study revealed several notable patterns. Statistical analysis showed that 35 pieces (approximately 49%) focused on political parties, making it the most represented category. This was followed by content concerning the Army and Police, which accounted for around 31% of the total. About 17% of the content was related to the Interim Government, while issues surrounding the July Mass Uprising and its leaders appeared in approximately 11% of the videos.
The dissemination of AI-generated content was also found to increase during accidents and emergency events. A significant example during the study period was the plane crash near Milestone School and College in Uttara, which drew considerable online attention. Approximately 10% of the analyzed content was associated with such accident-related incidents.
Additionally, around 10% of the sampled materials were classified as gender-insensitive content.
Table 2- Thematic Trends in the Content Included in the Research
Framing: Positive vs. Negative Narratives
Political Parties
Although 49% of the samples contained content related to political parties, this material was primarily concentrated on four major parties. Among these, 68% of the content focused on the Awami League, followed by 17% on the BNP, 9% on Jamaat-e-Islami, and 6% on the NCP.
Notably, nearly all AI-generated content concerning the Awami League portrayed the party in a favorable manner, reflecting a clear trend of positive propaganda aimed at enhancing the party’s image and political standing.
AI-generated content targeting the security forces was predominantly negative in tone, often portraying these forces unfavorably. The contents were frequently linked to political parties and events surrounding the July Mass Uprising.
For instance, in August 2025, an image circulated online showing several students in school uniforms standing in formation, with the front student holding a placard stating: “We want Sheikh Hasina.” Upon verification, this image was found to be inauthentic and generated using AI.
Analysis of contents related to political parties, the Interim Government, and the July Mass Uprising revealed that negative propaganda targeted all entities except the Awami League.
Security Forces (Army and Police)
Similarly, content concerning the Police and Army was largely negative, frequently associating these security forces with various political parties and the events of the July Mass Uprising. For instance, in September 2025, a video circulated online claiming to show the public chanting slogans against Army Chief Waker-uz-Zaman. Upon verification, the video was confirmed to be a fabricated deepfake created using AI.
Interim Government
A similar pattern of negative campaigning was observed targeting the Interim Government, including its Chief Adviser and other advisers. Of the analyzed samples, all 12 content pieces (approximately 17%) related to the Interim Government were negative in tone.
For example, in July 2025, a video titled “The Real Face of the Government Exposed” circulated widely online. In the video, an elderly man is heard saying: “Remove the Yunus government, save the country. This government is not undertaking any development or reforms for the poor.” Verification confirmed that the video was inauthentic and produced using AI.
Similarly, in August 2025, a video circulated online falsely claiming to feature a speech in the Indian Parliament referring to fascist Sheikh Hasina. In the video, a female speaker is purportedly heard commenting on replacing Dr. Muhammad Yunus with Sheikh Hasina. Verification confirmed that this video was created using AI.
Additionally, in July 2025, another video was disseminated online claiming that the Financial Adviser had recommended investing funds in a government investment platform with guaranteed significant weekly returns. This video was also fabricated through AI-driven digital manipulation.
July Uprising
Eight content pieces (approximately 11% of the sample) focused on the July Mass Uprising. A clear pattern of negative campaigning was evident, as all eight pieces presented adverse narratives about the event.
For example, in August 2025, a video circulated claiming that IGP Baharul Alam had declared August 5th as Police Martyrdom Day, honored the fallen, and promised justice. Verification revealed that the video was fabricated using AI technology.
Similarly, in July 2025, an interview video featuring an elderly person circulated online. When asked about August 5th, the individual was quoted as saying, “Ignorant students sold out for money and destroyed the country.” This video was also confirmed to be AI-generated.
Disasters and Emergencies
Fake news was observed to intensify during national disasters, with several AI-generated content pieces widely disseminated online during the tragedy. For instance, following the plane crash near Milestone School and College in Uttara, Diabari, Dhaka, an AI-generated image circulated on social media, falsely claiming to depict the scene of the aircraft wreckage.
Gender-Insensitive Content
Approximately 10% of the analyzed content contained gender-insensitive subject matter, often targeting female politicians or advisers of the Interim Government. For example, in August 2025, an image circulated online falsely showing Asif Mahmud Sajib Bhuiyan, the Interim Government’s Adviser for Youth, Sports, and Labour and Employment, with a young woman in a hotel room; the image was later confirmed to be AI-generated.
Similarly, in July 2025, an image of Dr. Tasnim Jara, Senior Joint Member Secretary of the National Citizens' Party (NCP), appearing in half pants went viral. This image was fabricated by editing an authentic photograph using AI technology. In September 2025, two images were circulated online, falsely claimed to be private photos of Fatima Tasnim Juma, Liberation War and Democratic Movement Secretary of DUSCU. Investigation revealed that these images were fabricated by superimposing Juma’s face onto pictures of two different women using AI.
AI-generated content was also observed on religious and other miscellaneous issues, although the prevalence of such content remained low during the three-month study period.
Conclusion
The study clearly indicates that AI-generated image/video is most prevalent in content related to political parties. Analysis of data from the three-month period in 2025 reveals a disproportionate prevalence of negative AI-generated disinformation targeting the Interim Government, the Police/Army, and political parties other than the Awami League. The only observed positive campaigning was directed exclusively toward the Awami League, with virtually all AI-generated propaganda exhibiting a bias in favor of the party.
Additionally, there is a clear trend of negative campaigning surrounding the July Mass Uprising, with all related content portraying the uprising and its participants in a criminalized and unfavorable light.
Research Limitations
The primary limitation of this study is its temporal scope, as it covers only a three-month period, preventing a comprehensive assessment of long-term trends and the evolution of AI-generated content. Another significant challenge was the exclusion of AI-generated videos whose links were no longer functional or had been removed due to platform policy violations. This limitation affected the completeness of the sample, and the characteristics of the excluded content may have potentially influenced the overall findings.
[''AI assistance was utilized for translation and editorial refinement of this report.']
FullBangla Post Link https://banglafact.com/postDetails/319Topics:
যুবদল নেতা হত্যার ঘটনাকে ভারতীয় মিডিয়ায় হাসিনার রায় পরবর্তী সহিংসতা হিসেবে প্রচার
Indian media frames the murder of a Jubo Dal leader as post-verdict violence following Hasina’s sentencing
ইন্টারনেটে আগুন সন্ত্রাসের নির্দেশনা ও উষ্কানি দিচ্ছে আওয়ামী লীগের এক্টিভিস্টরা
বিভ্রান্তি ছড়াচ্ছে পাসপোর্ট সূচকের খবর
গাজীপুরে নিহত সাংবাদিক তুহিন কি আদৌ ‘চাঁদাবাজি’ নিয়ে কোনো লাইভ করেছিলেন?
আপনার মতামত দিন
এই পোস্টটি কি আপনার জন্য সহায়ক ছিল?
এখনো কেউ ভোট দেয়নি। আপনিই প্রথম হোন!
0%
0%
আপনার মতামত শেয়ার করুন:
| মন্তব্য সমূহ:
এখনও কোনো মন্তব্য নেই। প্রথম মন্তব্যটি করুন!
| আরও পড়ুন
.jpg)
ইউক্রেন যুদ্ধকে ঘিরে জটিলতার কারণে এশিয়ার পাঁচ দেশে সফর স্থগিত করেছেন ইতালীর প্রধানমন্ত্রী, কিন্তু শুধু বাংলাদেশের নাম উল্লেখ করে শিরোনাম করেছে কয়েকটি গণমাধ্যম।
10 Rumors That Fueled the Tension in Khagrachhari
.jpg)
প্রতিষ্ঠিত গণমাধ্যম যখন বিভ্রান্তির উৎস
.png)
দিনাজপুরে ভবেশ চন্দ্র রায়ের মৃত্যু: বাংলাদেশের গণমাধ্যমে ভিন্ন ভিন্ন তথ্য প্রচার
বিশ্লেষণ
AI-Generated Online Content
Targeting Political Parties, Government, and Security Forces
২৮ অক্টোবর ২০২৫
Artificial intelligence (AI) has emerged as a significant tool in the domain of information warfare, notably intensifying the capabilities of propagandists. The accelerated generation and propagation of content pertaining to politically controversial and high-stakes social topics is demonstrably affecting public discourse.
The strategic deployment of propaganda is particularly potent in creating societal polarization and shaping public opinion, especially during critical windows such as pre-election cycles, periods of civil unrest, and emergency periods.
AI technologies are being intentionally leveraged for targeted harm, including systematic campaigns of defamation and digital abuse. A related and growing concern is the systematic utilization of AI to proliferate gender-insensitive and misogynistic content across various digital and social media ecosystems.
The Fact-Check and Media Research Team, BanglaFact, of the Press Institute of Bangladesh (PIB) conducted a study of AI-generated online content concerning Bangladeshi issues circulated across social media platforms between July and September 2025. The objective of the study was to identify the dominant themes and political biases embedded in AI-generated videos, with a view to assessing the scope, influence, and potential risks of AI-driven propaganda in Bangladesh.
The report revealed that the majority of AI-generated content—particularly forms of mis(dis)information—targeted political parties, the government, and security forces (police and army). Among the political parties, content related to the Awami League appeared most frequently; however, these materials primarily functioned as positive propaganda in favor of the party, whereas content concerning the government and other political parties was predominantly negative in tone and framing.
Research Methodology
A total of 71 AI-generated content pieces were selected as samples for this analysis. The samples were chosen using a purposive sampling method, focusing on videos considered likely to exert political or social influence across social media platforms. The selection process primarily involved reviewing fact-check reports published by various Bangladeshi fact-checking organizations during the study period (July–September 2025). Additionally, AI-generated videos disseminated through different media outlets within this timeframe were identified using relevant keyword searches.
Content was collected from Facebook, YouTube, Instagram, and X (formerly Twitter). In many cases, the same videos were found circulating across multiple platforms, and several had been fact-checked by more than one organization. The verification of samples utilized a range of AI-detection and media authentication tools, including but not limited to Deepfake-O-Meter, Hive Moderation, Sightengine, Cantilux, and Resemble AI Detect.
Analysis and Description
A total of 71 AI-generated content pieces were analyzed for this study, with each video or image treated as a distinct unit of analysis. The samples were categorized into the following thematic groups: Awami League, BNP, Jamaat-e-Islami, NCP, Interim Government, Army/Police, July Mass Uprising, Accidents, Religious, Gender-Insensitive, DUSCU (Dhaka University Central Students’ Union), and Others. These categories were not mutually exclusive, as a single video could be classified under multiple themes. All percentages were calculated based on the total number of analyzed images/videos (N = 71).
Platform and Medium
The study found that most AI-generated fake news circulated across Facebook, X (formerly Twitter), TikTok, YouTube, Instagram, and Threads. Several content pieces appeared simultaneously on multiple platforms, highlighting the coordinated and multi-platform nature of the disinformation campaigns.
Among these platforms, Facebook exhibited the highest prevalence, with 61 content pieces (86%), followed by TikTok, where 40 content pieces (approximately 56%) were identified. In contrast, X, Threads, Instagram, and YouTube showed a comparatively lower presence of such AI-generated materials.
Table 1- The presence of AI-generated misinformation on social media platforms.
Content Formats
According to the content analysis, video emerged as the most common format for employing AI technology to disseminate propaganda or misinformation. Of the 71 analyzed content pieces, 57 were video-based, while 14 consisted primarily of images.
Overview of Content Categories
The examination of the 71 AI-generated content pieces included in this study revealed several notable patterns. Statistical analysis showed that 35 pieces (approximately 49%) focused on political parties, making it the most represented category. This was followed by content concerning the Army and Police, which accounted for around 31% of the total. About 17% of the content was related to the Interim Government, while issues surrounding the July Mass Uprising and its leaders appeared in approximately 11% of the videos.
The dissemination of AI-generated content was also found to increase during accidents and emergency events. A significant example during the study period was the plane crash near Milestone School and College in Uttara, which drew considerable online attention. Approximately 10% of the analyzed content was associated with such accident-related incidents.
Additionally, around 10% of the sampled materials were classified as gender-insensitive content.
Table 2- Thematic Trends in the Content Included in the Research
Framing: Positive vs. Negative Narratives
Political Parties
Although 49% of the samples contained content related to political parties, this material was primarily concentrated on four major parties. Among these, 68% of the content focused on the Awami League, followed by 17% on the BNP, 9% on Jamaat-e-Islami, and 6% on the NCP.
Notably, nearly all AI-generated content concerning the Awami League portrayed the party in a favorable manner, reflecting a clear trend of positive propaganda aimed at enhancing the party’s image and political standing.
AI-generated content targeting the security forces was predominantly negative in tone, often portraying these forces unfavorably. The contents were frequently linked to political parties and events surrounding the July Mass Uprising.
For instance, in August 2025, an image circulated online showing several students in school uniforms standing in formation, with the front student holding a placard stating: “We want Sheikh Hasina.” Upon verification, this image was found to be inauthentic and generated using AI.
Analysis of contents related to political parties, the Interim Government, and the July Mass Uprising revealed that negative propaganda targeted all entities except the Awami League.
Security Forces (Army and Police)
Similarly, content concerning the Police and Army was largely negative, frequently associating these security forces with various political parties and the events of the July Mass Uprising. For instance, in September 2025, a video circulated online claiming to show the public chanting slogans against Army Chief Waker-uz-Zaman. Upon verification, the video was confirmed to be a fabricated deepfake created using AI.
Interim Government
A similar pattern of negative campaigning was observed targeting the Interim Government, including its Chief Adviser and other advisers. Of the analyzed samples, all 12 content pieces (approximately 17%) related to the Interim Government were negative in tone.
For example, in July 2025, a video titled “The Real Face of the Government Exposed” circulated widely online. In the video, an elderly man is heard saying: “Remove the Yunus government, save the country. This government is not undertaking any development or reforms for the poor.” Verification confirmed that the video was inauthentic and produced using AI.
Similarly, in August 2025, a video circulated online falsely claiming to feature a speech in the Indian Parliament referring to fascist Sheikh Hasina. In the video, a female speaker is purportedly heard commenting on replacing Dr. Muhammad Yunus with Sheikh Hasina. Verification confirmed that this video was created using AI.
Additionally, in July 2025, another video was disseminated online claiming that the Financial Adviser had recommended investing funds in a government investment platform with guaranteed significant weekly returns. This video was also fabricated through AI-driven digital manipulation.
July Uprising
Eight content pieces (approximately 11% of the sample) focused on the July Mass Uprising. A clear pattern of negative campaigning was evident, as all eight pieces presented adverse narratives about the event.
For example, in August 2025, a video circulated claiming that IGP Baharul Alam had declared August 5th as Police Martyrdom Day, honored the fallen, and promised justice. Verification revealed that the video was fabricated using AI technology.
Similarly, in July 2025, an interview video featuring an elderly person circulated online. When asked about August 5th, the individual was quoted as saying, “Ignorant students sold out for money and destroyed the country.” This video was also confirmed to be AI-generated.
Disasters and Emergencies
Fake news was observed to intensify during national disasters, with several AI-generated content pieces widely disseminated online during the tragedy. For instance, following the plane crash near Milestone School and College in Uttara, Diabari, Dhaka, an AI-generated image circulated on social media, falsely claiming to depict the scene of the aircraft wreckage.
Gender-Insensitive Content
Approximately 10% of the analyzed content contained gender-insensitive subject matter, often targeting female politicians or advisers of the Interim Government. For example, in August 2025, an image circulated online falsely showing Asif Mahmud Sajib Bhuiyan, the Interim Government’s Adviser for Youth, Sports, and Labour and Employment, with a young woman in a hotel room; the image was later confirmed to be AI-generated.
Similarly, in July 2025, an image of Dr. Tasnim Jara, Senior Joint Member Secretary of the National Citizens' Party (NCP), appearing in half pants went viral. This image was fabricated by editing an authentic photograph using AI technology. In September 2025, two images were circulated online, falsely claimed to be private photos of Fatima Tasnim Juma, Liberation War and Democratic Movement Secretary of DUSCU. Investigation revealed that these images were fabricated by superimposing Juma’s face onto pictures of two different women using AI.
AI-generated content was also observed on religious and other miscellaneous issues, although the prevalence of such content remained low during the three-month study period.
Conclusion
The study clearly indicates that AI-generated image/video is most prevalent in content related to political parties. Analysis of data from the three-month period in 2025 reveals a disproportionate prevalence of negative AI-generated disinformation targeting the Interim Government, the Police/Army, and political parties other than the Awami League. The only observed positive campaigning was directed exclusively toward the Awami League, with virtually all AI-generated propaganda exhibiting a bias in favor of the party.
Additionally, there is a clear trend of negative campaigning surrounding the July Mass Uprising, with all related content portraying the uprising and its participants in a criminalized and unfavorable light.
Research Limitations
The primary limitation of this study is its temporal scope, as it covers only a three-month period, preventing a comprehensive assessment of long-term trends and the evolution of AI-generated content. Another significant challenge was the exclusion of AI-generated videos whose links were no longer functional or had been removed due to platform policy violations. This limitation affected the completeness of the sample, and the characteristics of the excluded content may have potentially influenced the overall findings.
[''AI assistance was utilized for translation and editorial refinement of this report.']
FullBangla Post Link https://banglafact.com/postDetails/319