From Digital Divide to ‘AI Poverty’: Bangladesh’s Next Development Challenge

The Calculus of a New Divide: Artificial Intelligence (AI) is no longer a distant futuristic concept for Bangladesh. It has quietly integrated into our classrooms, banks, hospitals, media houses, call centers, government offices, agricultural services, and small businesses. For some, it promises expedited services, new employment opportunities, heightened productivity, and creative avenues. For others—particularly the impoverished, rural populations, women, persons with disabilities, linguistic minorities, and informal sector workers—it threatens to become a formidable, invisible wall of exclusion.

This emerging socio-economic disparity can be aptly termed "AI Poverty."

AI Poverty is not merely the inability to access ChatGPT or utilize a specific software tool; its implications run much deeper. It refers to the systemic exclusion of marginalized individuals from the data ecosystems, technical skills, languages, infrastructure, institutional capacities, and decision-making processes that power AI. In this setup, people may remain connected but fundamentally disempowered. They are counted as data points, but their voices remain unheard. They are subjected to algorithmic systems without the agency to question how those systems operate.

Bangladesh has already witnessed an earlier iteration of this challenge: the Digital Divide. For decades, national discourse centered around mobile connectivity, internet penetration, and digital financial inclusion. The country’s achievements here are undeniable. By March 2026, mobile subscriptions surpassed 186 million, and internet connections reached nearly 130 million. For a nation of roughly 175 million people, this represents commendable progress.

The Illusion of Numbers vs. True Inclusion: However, raw statistics often mask the true state of inclusion. A mobile connection or a single smartphone in a household does not guarantee that women, children, the elderly, or informal laborers can navigate digital spaces safely, confidently, and productively. Connectivity is merely the threshold. The critical question remains: Who can walk through that door, understand the digital landscape, and leverage it to improve their socio-economic standing?

This inquiry becomes highly critical as the United Nations hosts the Global Dialogue on AI Governance in Geneva in July 2026. This dialogue aims to unite governments, the private sector, academia, the tech community, and civil society to forge an international framework for AI governance. The core agenda spans AI’s societal impacts, bridging the tech divide, ensuring safe and trustworthy AI, protecting human rights, and establishing accountability.

For Bangladesh, this is not just a distant diplomatic exercise; it is a core development challenge. If global AI governance is shaped solely by powerful economies and tech conglomerates, developing countries like Bangladesh will remain trapped in a cycle of digital dependency—using systems built by others, relying on models trained on foreign data, and operating under rules written without our input. That is not digital transformation; it is digital dependency.

Connectivity Alone Cannot Mitigate AI Poverty: The Bangladesh Bureau of Statistics (BBS) ICT Access and Use Survey 2024-25 offers a sobering reality check. While individual internet penetration stands at 53.4% nationwide, the urban-rural disparity remains stark—75.7% in urban areas compared to just 43.6% in rural regions, marking a 32.1 percentage point gap. Gender disparities are equally pronounced, with men outnumbering women in both internet usage and smartphone ownership.

These metrics matter because AI does not thrive on political rhetoric. It is built upon data, devices, language, human skillsets, and institutional frameworks.

A rural student with an internet connection still misses out on the benefits of AI without trained teachers, quality devices, and localized educational content. Similarly, a smallholder farmer lacking a digital cultivation history or formal financial footprints risks becoming completely invisible to AI-driven credit scoring systems. Furthermore, without robust digital safeguards, women face heightened vulnerabilities to AI-generated harms like deepfakes and targeted cyber-harassment.

The greatest misconception in technology policy is conflating tool accessibility with genuine inclusion. Installing a chatbot in a school does not reform education if teachers are untrained and the content is detached from the curriculum. Implementing AI in banking does not foster financial inclusion if the poor have no historical data footprint. An AI health tool cannot improve public health if rural medical records are missing and women's health realities are omitted from datasets.

Data is inherently subject to human bias; it reflects only those who were counted. If Bangladesh’s AI ecosystem is trained predominantly on urban, institutional, English-centric, or elite datasets, it will amplify the visibility of the privileged while rendering the marginalized even more invisible.

While the traditional Digital Divide separated those with internet access from those without, the AI Divide separates those who own, control, and profit from data from those who are unconsciously governed by its algorithms. Against the backdrop of current economic pressures, high inflation, and strained household incomes, the recurring costs of data, device maintenance, and digital security create a trap of "partial inclusion"—where citizens exist online but cannot extract meaningful value from the ecosystem.

From Framework to Field Execution: Bangladesh’s draft National Artificial Intelligence Policy 2026-2030 rightly identifies several imperative areas: risk-based regulations, algorithmic impact assessments, data governance, and the necessity of human oversight. These are essential foundational steps, but the true test lies in execution.

AI governance must not devolve into exclusive seminars confined to Dhaka-based bureaucrats, tech executives, and development donors. If AI is to serve the public good, the policy-making process must actively engage rural educators, community healthcare workers, women's rights organizations, farmers, journalists, and youth networks. We must avoid past mistakes: building systems for the people without designing them with the people.

The Way Forward: A Human-Centric AI Agenda: To ensure AI democratizes power rather than centralizing it, Bangladesh must prioritize the following strategic pillars: 

Data Justice as Public Infrastructure: Just like roads and fiber-optic cables, high-quality, secure, and representative datasets in health, education, and agriculture must be treated as essential public goods. Data governance must uphold informed consent, privacy, and community benefit, preventing citizens from being treated merely as raw material for data extraction.

Civic AI Literacy: AI literacy must transcend basic coding skills. Citizens must be equipped to understand how algorithms function, identify systemic biases, spot deepfakes, and know how to appeal automated decisions that impact their rights.

• Investment in Localized and Vernacular AI: Substantial public and private investment must be directed toward AI models that comprehend the Bengali language, regional dialects, and indigenous languages, ensuring localized advisory content for agriculture, disaster preparedness, and legal aid.

• Labor Protection Against Blind Automation: Given our economic reliance on ready-made garments, agriculture, and informal labor, sector-specific labor impact assessments are crucial to ensure AI acts as a tool for elevating productivity, wages, and dignity rather than silently replacing human workers.

• Algorithmic Accountability: Before deploying AI in public service delivery, systems must undergo independent audits to verify data integrity, inclusivity, and linguistic accessibility, ensuring clear mechanisms for citizen grievance redressal.

• Leveraging Community Institutions: Bangladesh's historic development milestones were achieved through grassroots mobilization. Integrating community radios, local civil society, public libraries, and youth clubs will help translate complex technological concepts into tangible public assets.

On the global stage, Bangladesh must advocate clearly that bridging the AI divide requires more than just application access; it requires technology transfers, open-source resources, affordable infrastructure, and the meaningful representation of the Global South.

If AI empowers a rural woman to access critical health advice in her native dialect, shields her from online exploitation, and bolsters her livelihood, the technology has succeeded. If it merely automates systemic exclusion and concentrates wealth among the tech-elite, it has failed. AI poverty is preventable, but avoiding it demands a shift away from market-centric approaches toward a human-first roadmap. The digital era promised to leave no one behind; the AI era will test the truth of that promise.

• A. H. M. Bazlur Rahman: Expert in digital democracy, information integrity, and media development policy; Ambassador for Responsible Artificial Intelligence and Antimicrobial Resistance (AMR).