An Analysis on the National Artificial Intelligence Policy 2026-2030 (Under Consultation) towards countering AI-Driven Electoral Disinformation in Local Governments of Bangladesh
PRELUDE:
The cognition bias in the digital world is equipped with the insisted frame of Internet and Algorithm with continuation of privileged actors who exert the soft power only to influence the Internet mechanism on impressing the narrative in the digital space. This cognition bias is known as algorithmic bias which has a role to influence the voters in this current democratic backsliding. Algorithmic is a means through which a computer program sets a combination of instructions and training mechanisms to abide by certain tasks to solve a problem. Or in simple words, it is a trained mechanism to function what to do and how to do. The functions can transform the input of the data through processing it to useful output of the problem. Even in case of analyzing the input data, interface can also provide results on the instructions and allow decisions on its behalf.
The 2026 social media algorithm ingests and understands data through a sophisticated “Intelligence Core” that employs advanced AI technologies. This core acts as the primary data ingestion system, meticulously analyzing every piece of content and user interaction. It moves beyond simple metadata, using deep learning to interpret visual, textual, and behavioral cues. This comprehensive data intake allows the algorithm to precisely categorize content, understand user intent, and detect sentiment, forming the foundational layer for content scoring and distribution. By processing vast amounts of diverse data, the algorithm builds a rich, nuanced profile of both content and audience. Source: https://mindmapai.app/mind-mapping/the-2026-social-media-algorithm
As in Bangladesh, people use Facebook mostly as data speaks Facebook had 75.18 million users in Bangladesh in August 2026, equivalent to 41.5% of the country’s total population. Men accounted for 62.6% of the user base, while people aged 18–24 formed the largest group with 29.6 million users. The widest gender gap appeared among users aged 25–34, where men outnumbered women by 18.4 million. The parent company of Facebook, Meta, creates a balance of engagement with AI in signaling social relationships meaning it forms a complex equation that the behavioral pattern in TikTok. Facebook deeply analyses how the user’s interaction with someone was before and how relevant their content is to their interests and the likelihood that the user will engage or not. During the elections, voters tend to fall into this interactive model of algorithm. Therefore, quality content made by AI has rewards in built in it through reels and often posting these ensures reach with genuine engagement through fear mongering, backlashes and more. Facebook as a platform does not necessarily endorse AI generated disinformation but system inherently amplifies the AI generated content due to its engagement, sensationalist and emotionally charged factors during the elections.
LITERATURE REVIEW:
In his paper, Gaurav Goswami pointed out AI-driven recommendation systems have significantly influenced digital information consumption, leading to the formation of “AI echo chambers.” These chambers reinforced user preexisting perspective through feeding similar contents which exposed them towards diversity. In the case of US elections in 2016, Goswami examined possible solutions on few aspects of the issue:
- Algorithmic transparency
- Content Diversification
- Regulatory measures
- AI Ethical Framework
In Bangladesh, I am also highlighting the political party’s responsiveness as an urgency to ensure AI governance to mitigate misinformation, radicalization and digital polarization between contested candidates. In quest to holding accountable to the algorithms through acknowledging the fact of upregulations has caused serious polarization of the parties. These platforms are private entities which present algorithms as a bug, but these issues do not have proper fixations. Therefore, such discrimination and accountability are not just limited to the interconnected network but also seen in law enforcement algorithms known for forecasts. The algorithm can be based on stereotypical biases such as foreign origin and ‘foreign sounding’ names, forcing them to start distortions. The dilemma starts when this system decides which candidate should be endorsed on the front tab, which person to be highlighted and whose content should be censored. When private actors are making the claim on the elimination of the algorithmic bias, the progress towards achieving that hinders and thus the policies do not see the light. Bangladesh needs an intersectional approach to aware prejudice and biases produced through algorithmic dilemma and encourage the inspections by the state authorities and election commission on the interpretation upon the voters. The work states that responders depicted that AI generated misinformation were influential in the outcome of the presidential elections in a broader public perspective. The use case of AI did not increase the trust among the voters, but it surely revealed concern of mistrust, fear, lack of transparency and institutional reactions over the legitimacy of the process.
Participants perceive that AI-generated disinformation influenced the outcome of the presidential election. There is broad public support for regulating the use of AI in election campaigns, although debates persist about regulatory mechanisms and models. The use of AI did not increase trust in candidates or in the legitimacy of the electoral process. Sentiment analysis reveals widespread concern, mistrust, and fear, along with a demand for transparency, digital literacy, and institutional action. The author mentioned that the solution does not rely on ban, but it should have ethical and transparent regulations. In case of Ecuador, author mentioned about a regulatory framework on the National Election Council to label obligations, penalization and verification of the mechanisms. It has also been stated that media literacy policies also enforce a structural risk towards democracy beside the algorithmic dilemma in multiple actors to ensure free, fair and informed elections. AI has been seen as a structural risk to democratic practices by facilitating manipulation on massive scale and with high degree of realism.
Based on the data collection, the people of Ecuador agree on the regulation of generative AI in the electoral campaigns, but the mechanisms and model need to be discussed and suggested towards Consejo Nacional Electoral. The work does not focus on the responsiveness of the political parties on how they perceived this whole issue that the use of AI in Ecuador’s presidential campaign in 2025 did not increase trust in the candidates and the outcome of the election of the President of the Republic. This article focuses on the direct challenges towards democratic process due to amplified falsehood distorted in the political environment through algorithms. This work aims to analyze the role of AI in disinformation campaigns and assess the limitations of the AI framework on governance. It focuses that the urgent demand of the AI based approaches spearhead the democratic governance mechanisms. It also refers to a multi stake holder approach that will involve the accountability, enforceability of regulations and sustainable civic education on the social media platforms against democratic resilience.
In case of India, the report states that the policymakers have approached disinformation in its amendments to IT regulations. In 2023, India introduced further measures to amend IT rules to create government backed fact checking unit to identify the false information regarding governance-oriented information. But India is yet to provide same artifact in its democratic processes as platforms like Twitter, Facebook, and WhatsApp are now required to comply with official takedown requests targeting content deemed “fake news” by authorities (Mehrotra and Upadhyay, 2025). The article states that in this evolving contestation of truth and false, AI is weaponized and acts as a shield through its deployment in the application and ecosystem. But our research focuses on the weaponized part where AI’s use case is assessed on the electoral disinformation and how political parties are responding during an election campaign. In the evolving contest between disinformation and truth, AI is both a weapon and a shield. Its deployment must be grounded in an understanding that harm stems not from the existence of a tool, but from its application and the ecosystem of incentives surrounding it. Therefore, any policy framework must prioritize intent-based regulation, investment in cross-sector R&D, and transparency-by-design. As the information environment grows more complex, building a layered, adaptive, and evidence-based response that leverages the full spectrum of AI capabilities is the only viable path forward. This review underscores that safeguarding our digital public sphere is not a matter of banning technology, but of channeling it toward resilience, accountability, and collective intelligence.
LOCAL GOVERNMENT ELECTION OF BANGLADESH:
Bangladesh is set for its next Local Government Election in October which is based on a non-partisan basis. Interim Government has discharged ordinances to formulate the functionality of the local government representatives as included on City Corporations, Municipality Mayors and Councilors across the country. Bangladesh Election Commission has drafted Code of Conduct on all the tiers as it proposes strict regulations on the use of sound system and campaign hours. But have there been any regulations on the use of AI in electoral disinformation?
Bangladesh is drafting a National Artificial Intelligence Policy grounded in human rights, ethics and responsible innovation, Posts, Telecommunications and Information Technology Minister Faqir Mahbub Anam announced at the World Artificial Intelligence Conference in Shanghai. Bangladesh has also drafted a revised version of National Artificial Intelligence Policy 2026-2030 by the ICT Division for public review. It proposes a risk-based approach to AI regulation with prohibitions on certain uses, impact assessments for high-risk systems, and institutional arrangements for coordination and capacity building.
POLICY ANALYSIS:
In Section 4.1, the government shall prohibit AI-enabled election interference, including creating or spreading AI-generated disinformation, deepfakes, or targeted manipulation meant to undermine electoral integrity or democratic participation. In Section 5.5, the policy also calls for strict boundaries on AI use case in campaigning, micro targeting, voter profiling and manipulation of public opinion in quest to protect democratic integrity. Now as per the draft, the manipulation terminology is not defined as it states in 5.5 that it shall define strict boundaries, but the policy itself does not coin the term or set regulations. We can see that it does not formulate legislation, rather an aspirational term which cannot be enforceable.
In 4.1, it has been stated that disinformation is banned but instead of any further consequences the policy has not formulated any monitoring cell or reporting mechanism during election campaigns. Therefore, election disinformation does not have any complaining mechanism through strict monitoring. In 4.5, it has been stated that Government Agencies cannot label political speech, criticism, satire or journalism as AI Harmful Content. This should have been a protection towards free speech, but it is drafted in the Second Revision as a Gendered-harm Content. In this case, any party can term AI generated disinformation as a Satire or Political Commentary to resist the complaint. This National Artificial Intelligence Policy has no Election Oriented Regime based on polling days, campaign period. AI Oriented Electoral Disinformation has been termed into general political activities which have no legal scrutiny when it matters. In 4.5, there lies a scope for Labeling AI content and to remove it within 24 hours as Harmful AI Content, but it does not operate on a similar basis to political deepfakes and disinformation as per the policy.
CONCLUSION:
It has been remarkable to see Bangladesh working on the policy level on Artificial Intelligence. But it should prioritize the Digital Peacebuilding aspect of the policy which refers to the use of technology to function conflict analysis, violence prevention and foster peace in the post conflict situation. After the fall of the regime on 5th August, Bangladesh is set towards post conflict peacebuilding so that institutions do not incur to such atrocities again and no other states person can build autocratic behavior through democratic backsliding. The goal is to bring people together through stopping Hate Speech. Disinformation and Digital Division. During July 2024, we witnessed how government suppressed Internet Blackout and focused on Digital Spy to stop the protests . The principles do state that AI should not intervene in the election campaigns, but the mechanisms and tools are still not drafted in the policy.
REFERENCE:
- Dhaka Tribune. (2026, July 18). ICT minister: Bangladesh drafting human-centric national AI policy. https://www.dhakatribune.com/bangladesh/415452/ict-minister-bangladesh-drafting-human-centric
- Goswami, G. (2025, February 26). AI echo chambers: How algorithms shape reality, influence democracy, and fuel misinformation. TechRxiv. https://doi.org/10.36227/techrxiv.174059950.03385147/v1
- ICT Division, Ministry of Posts, Telecommunications and Information Technology, Government of the People’s Republic of Bangladesh. (2026, February 9). National Artificial Intelligence Policy 2026–2030 (Draft v2) [Unpublished draft policy document].
- Mehrotra, A., & Upadhyay, A. (2025, March 19). India’s growing misinformation crisis a threat to democracy. Setopati. https://en.setopati.com/International/164366
- MindMapAI. (2026, March 13). The 2026 social media algorithm. https://mindmapai.app/mind-mapping/the-2026-social-media-algorithm
- NapoleonCat. (2026, August). Facebook users in Bangladesh – August 2026. https://stats.napoleoncat.com/facebook-users-in-bangladesh/
- Romanishyn, A., Malytska, O., & Goncharuk, V. (2025). AI-driven disinformation: Policy recommendations for democratic resilience. Frontiers in Artificial Intelligence, 8, Article 1569115. https://doi.org/10.3389/frai.2025.1569115
- Suing, A. (2025). Disinformation, AI and regulation in Ecuador’s 2025 presidential election. Frontiers in Political Science, 7, Article 1624206. https://doi.org/10.3389/fpos.2025.1624206
About Author:
Syed Al Farabi
Undergraduate student,
Department of peace and conflict studies,
University of Dhaka
AUTHOR NOTE:
Mr Syed Al Farabi is a junior undergraduate student in the Department of Peace and Conflict Studies at the University of Dhaka. As a Master Trainer of Applied Democracy Lab, he focuses on Democratic backsliding in the world and how digital peacebuilding can help a state towards building strong institutions.