Bangladesh’s AI Policy Needs an Engine, Not Just a Map

This is the third article in a four-part series that began with Bangladesh’s AI Moment and What Bangladesh’s AI Policy Must Get Right. Those earlier pieces explored the vision behind Bangladesh’s AI ambitions and the policy scaffolding needed to guide them.

After those articles were published, the UNESCO Bangladesh Artificial Intelligence Readiness Assessment Report 2025 arrived. It made the next steps inevitable. The report's diagnostic clarity showed not just what Bangladesh aspires to build, but what the logic demands: that vision and policy mean nothing without institutional infrastructure and inclusive participation.

We are trying to run an AI economy on institutional plumbing designed for a different era, and no amount of policy enthusiasm can compensate for the absence of a functioning operational engine.

A policy without execution mechanisms is not a plan. It is a press release.

The Governance Void

The report’s first significant finding is what I would call the missing middle of our AI governance stack. At present, responsibility is diffused across the ICT Division, the Cabinet, the Ministry of Law, and several others.

Collaboration sounds ideal, but in practice, fragmented responsibility often means no one is actually responsible.

When a ministerial AI pilot fails, who answers for it? When two agencies buy incompatible systems, who enforces standards?

UNESCO’s recommendation is clear. Bangladesh needs a Central Office for AI Governance, a single and empowered body that coordinates implementation, standards, interoperability, audits, and oversight.

Without this institutional anchor, our AI policies will oscillate between enthusiasm at the top and execution paralysis at the bottom.

In a related way, the report calls out another structural gap. Bangladesh still lacks an independent Data Protection and Cybersecurity Authority. The Personal Data Protection Ordinance 2025 is a step forward, but without an autonomous regulator to enforce it, it cannot build public trust, especially when AI systems increasingly make decisions with social and economic consequences.

The Procurement Black Hole

The government is the country’s largest buyer of digital systems, yet Bangladesh has no dedicated policy for procuring AI systems.

This is the operational bottleneck that could quietly derail the entire national AI agenda.

Traditional procurement rules reward the lowest bidder and the quickest timeline. But AI is not cement or office supplies. Deploying a flawed algorithm in healthcare, taxation, welfare, or law enforcement incurs more than just financial costs. It is human.

A flawed facial recognition system deployed in law enforcement doesn't just waste money; it generates false arrests. A biased welfare algorithm doesn't just underperform; it denies benefits to eligible families.

The UNESCO roadmap proposes a mandatory certification and audit program for AI vendors. This should not be optional. The state must ensure that any system it deploys must be able to:

●    Explain its decisions

●    Demonstrate bias testing

●    Pass security audits

●    Preserve human oversight

Otherwise, we risk importing systems we can neither inspect nor govern.

The Accountability Crisis

Bangladesh’s cybersecurity position is, at best, inconsistent. One global index ranks us mid-range. Another places us among the least secure nations in the world. Both assessments agree on one thing. We are not ready for the data and infrastructure needs demanded by an AI economy.

The new Cyber Security Ordinance 2025 is encouraging. It specifically names machine learning, machine vision, and large language models as digital systems subject to regulation.

But UNESCO is right to say the framework requires strengthening. If we are to build a national data backbone that powers AI systems, the security architecture around it cannot be porous.

Where We Go From Here

The roadmap identifies the institutional gaps: a central AI office, operational steering committees, certification mechanisms, procurement reform, and accountability frameworks. Those are the right ingredients.

But listing them is not the same as building them.

Establishing a central office requires budget allocation, ministerial authority, and staffing decisions. Certification mechanisms require technical expertise that Bangladesh does not yet have at scale. Procurement reform requires changing entrenched bureaucratic incentives.

These are not insurmountable problems. They are operational ones. And operational problems have operational solutions. But only if we are willing to invest in capacity, not just policy.

But even functional institutions are not enough. Infrastructure can execute. But who decides what gets executed, and for whom?

That question leads to the final piece of this arc: the people who will shape Bangladesh's AI future and the people at risk of being left behind.

Dr. Zunaid Kazi is an AI visionary and entrepreneur who has spent over 30 years turning complex ideas into intelligent systems