Bangladesh’s AI Moment: Testing the Implementation Gap
The Democratic Republic of the People of Bangladesh has drafted an AI Policy that takes governance seriously, aiming to address the significant challenges posed by artificial intelligence (AI) technology.
Overview of the National AI Policy 2026-2030
The National AI Policy 2026-2030, currently being finalized by the ICT Division, includes a risk-based regulatory framework, explicit prohibitions on mass surveillance and social scoring, mandatory algorithmic impact assessments for high-risk systems, and a commitment to ratify the Council of Europe’s Framework Convention on AI.
This policy arrives at a precarious moment, with the interim government led by Nobel laureate Muhammad Yunus preparing to transition power to an elected successor. The critical question is whether any government will build the necessary institutions to implement this policy.
Classification System and Prohibitions
The draft proposes a risk-based classification system that categorizes practices into four tiers, from prohibited practices to high-risk applications. The prohibited category includes:
- Real-time biometric surveillance in public spaces
- Social scoring systems
- AI-enabled weapons
High-risk classifications will trigger mandatory algorithmic impact assessments, human oversight requirements, and transparency obligations, marking a significant shift from previous governance approaches.
Historical Context of Digital Governance
Historically, Bangladesh has struggled with digital governance, where the previous government’s Digital Security Act was used to suppress free speech and silence dissent. The new policy seeks to build guardrails against digital authoritarianism, but its success hinges on the institutions enforcing it.
Institutional Framework and Implementation Challenges
The policy draws inspiration from the EU AI Act but is tailored to fit Bangladesh’s unique context, assigning the National Data Governance and Innovation Agency (NDGIA) as the coordinating body. However, the NDGIA is not yet operational, and the policy lacks a clear establishment timeline or capacity benchmarks. This raises concerns about implementation capabilities.
The policy also references an Independent Oversight Committee that requires legislative action, which has not yet been drafted.
AI Readiness Assessment
In November 2025, a collaborative AI Readiness Assessment Report was released by UNESCO, UNDP, and the ICT Division, identifying 15 priority actions and documenting significant gaps in Bangladesh’s AI landscape, such as:
- Fragmented data systems
- GPU scarcity
- Outdated curricula
- Lack of AI ethics instruction
- Gender disparities in the AI workforce
The new draft acknowledges these issues but fails to establish clear connections between the readiness report’s recommendations and the policy’s provisions.
The Language Infrastructure Challenge
A critical structural issue is the lack of language infrastructure for Bangla, which hinders the inclusive deployment of AI. Currently, only Hishab produces Bengali Large Language Models, a significant gap for a language spoken by over 170 million people. AI systems trained solely on English data may not adequately serve Bangladeshi needs.
Public service delivery, education, agriculture, and healthcare are sectors that could benefit greatly from AI, yet they require systems that operate effectively in Bangla.
Political Transition and Future Implications
As Bangladesh approaches its first election since the July Uprising, the continuity of this policy work amidst political transitions remains uncertain. Historical patterns suggest that political changes often reset technology governance agendas.
While the draft AI policy represents a significant step towards thoughtful governance, the gap between the policy’s language and actual institutional enforcement remains a concern. Sustained political commitment across electoral cycles will be essential for realizing the potential of AI governance in Bangladesh.
In conclusion, while Bangladesh has made strides in creating a framework for AI governance, the real challenge lies in building the necessary institutions, training regulators, and developing local AI infrastructure that reflects the unique needs of its population.
