How ready is Bangladesh for the era of AI?

This is the era of Artificial Intelligence, and Bangladesh is belatedly entering that era. But is the country's infrastructure ready for it?

It is against this backdrop that UNESCO in November launched the Readiness Assessment Methodology (RAM) report for Bangladesh. The standardised study was a controlled detonation to see whether Bangladesh's institutions could withstand even the smallest shock from emerging AI systems.

What emerged was a sobering reminder of how much catching up Bangladesh has to do. Every developing country has gaps, and in the case of Bangladesh, these gaps are convex: they widen precisely as intelligence scales. In simple words, intelligence is being layered on flimsy foundation.

RAM lays bare Bangladesh's AI Readiness Paradox with clinical clarity: stalled ordinances,policy drafts trapped in bureaucratic limbo, data scattered across ministerial silos, uneven infrastructure across districts and widening digital divides. 

The state appears digitally active but remains operationally inert. Anyone who has waited six months for a circular to move from desk to desk in Dhaka knows this is not a metaphor -- it is physics. 

The UK's 2020 A-Level algorithm scandal, still opaque in its mechanics, offers a glimpse into what institutional lag looks like when intelligence systems outpace oversight. 

Nearly 40 percent of students were downgraded before any authority could explain how the algorithm functioned, who controlled it or how to reverse its decisions. This was not a technical failure; it was a governance one.

 AI does not fail loudly at first -- it fails silently, at scale. It does not wait for literacy, legitimacy or leadership consensus. It expands while committees deliberate. Bangladesh is vulnerable precisely at this seam. Digital services have spread rapidly, but the institutional ballast needed to restrain AI remains thin -- like anchoring high-tension cables into soft alluvial soil.

The response to repeated public data leaks across NID, health and education systems provides the clearest example yet of the country's unpreparedness to the AI phenomenon. Has any institution meaningfully altered its processes, staffing or oversight? The answer, sadly, is no, according to RAM. The administrative machinery's behaviour remains unchanged -- same delays, same opacity, same weak enforcement. The threat is not artificial intelligence; the threat is artificial confidence.

True AI readiness is not a policy outcome; it is a property of system design. There must be room to fail safely, correct quickly and withdraw before damage metastasises. In resilient systems, pressure is a teacher; in brittle systems, it is an executioner.

Bangladesh's first fault line is data. RAM shows that the country performs poorly in availability, interoperability and open-data governance. 

AI feeds on dense, clean, contestable datasets -- not fragments hoarded across ministries without shared standards or consent mechanisms. When citizens become subjects of algorithmic decisions built on invisible data, intelligence turns extractive. Data poverty produces intelligence poverty -- every time.

The second fault line is skills. Bangladesh records more than one mobile subscription per person, yet less than half its population meaningfully uses the internet. This is not inclusion; it is surface penetration. RAM confirms what classrooms already reveal: AI education is sparse, urban-biased, disconnected from power reliability and basic infrastructure. 

The third fault line is governance. The AI policy is still a draft. Cybersecurity capacity remains shallow. Inter-ministerial coordination remains episodic. Most tellingly, RAM signals declining trust in public digital systems. Once lost, trust returns slowly, if at all.

The fourth fault line is culture. Intelligence imported without context becomes mimicry. Models trained on foreign languages, foreign datasets, foreign moral priors do not understand local economies, informal labour, minority communities or linguistic nuance. 

RAM notes Bangladesh's failure to build open Bangla datasets at scale, let alone those representing indigenous and marginal languages. Nations that do not encode themselves into AI will be interpreted by others -- inaccurately and permanently.

So what is the path forward? Not another policy. Not another strategy document. Not another task force. 

The solution is structural. Bangladesh must design AI systems that treat stress as information: systems that gain strength through stress rather than collapse under it.

That requires radical data transparency, a national AI sandbox, sovereign compute and cloud capacity, a language-first AI initiative and a citizen-first ethic.

Radical data transparency means every state dataset used by algorithms must be traceable, auditable, contestable and publicly legible. Opaque data is not a technical flaw; it is a political liability.

A national AI sandbox would entail controlled experimentation. Let failures be small, reversible and instructional. Progress comes from trial under constraint, not theory.

Without control over compute, Bangladesh will remain structurally dependent: consuming intelligence while exporting sovereignty. Digital dependence is geopolitical exposure.

Bangla and minority languages must be encoded deliberately, not left as afterthoughts. Intelligence that cannot speak the people's language will never serve them.

And AI must not harden existing asymmetries -- where the state scales upward and the citizen becomes smaller. The AI policy must be to ensure dignified citizenship.

The final truth is merciless: AI does not transform nations; it amplifies what is already there. Countries are rarely undone by technology itself -- they are undone by the illusions they build to avoid confronting their own fragility. In this new intelligence era, speed is not strategy. The winners will not be the nations that rushed to deploy, but the ones that learned to fail -- early, locally, reversibly -- before the stakes became existential.

The author is an assistant professor at North South University and member of UNESCO AI Ethics Experts Without Borders. He can be reached at zulkarin@gmail.com