Bangladesh - National AI Policy (2024)

Overview

The National Artificial Intelligence Policy 2024 (Draft) sets a national vision to integrate AI into Bangladesh's development agenda by promoting responsible, inclusive and human-centred AI. It is framed around six core principles (social equity; transparency and accountability; safety and robustness; sustainability; partnership and collaboration; human-centered AI) and identifies priority sectors for adoption including public services, healthcare, finance, agriculture, education, telecommunications and smart cities. The draft was published by the ICT Division in April 2024 and has been the subject of stakeholder consultations led by government departments and international partners (UNESCO/UNDP). The policy is a high-level instrument intended to be operationalized by subsequent laws, regulations and institutional mechanisms such as an independent National AI Centre of Excellence (NAICE) and a National AI Advisory Council. Key implementation building blocks include risk classification of AI systems, impact assessments, data governance safeguards and capacity building for research and skills development. For context and stakeholder analysis see coverage by national media and civil society reactions such as The Daily Star and The Business Standard.

Definitions

The draft provides working definitions for central terms used throughout the policy ("AI system", "model owner", "data controller/processor", "high-risk AI system", "automated decision-making", "human oversight"). Definitions adopt an outcomes-oriented approach: an "AI system" is any software that performs tasks commonly associated with human cognitive functions, including machine learning and rule-based algorithms. "High-risk AI system" is defined by impact on fundamental rights, critical public functions, safety or essential services. The draft emphasizes functional definitions to ensure the policy is technology-neutral and can accommodate evolving model types and deployment modes.

Governance and Institutional Framework

The draft places governance responsibility with the ICT Division (Ministry of Posts, Telecommunications and Information Technology) as policy lead while proposing the creation of a National Artificial Intelligence Centre of Excellence (NAICE) to coordinate implementation, standardization, certification and capacity building. A National AI Advisory Council (multistakeholder) is proposed to provide strategic oversight; membership examples include senior officials from the ICT Division, Posts & Telecommunications Division, Public Security Division, BTRC, National Cybersecurity Agency, a2i, industry bodies (BASIS, BACCO), and academia. The institutional design contemplates NAICE acting as a central registry hub, convenor for sectoral AI roadmaps, and technical secretariat for cross-ministerial coordination. Implementation responsibilities for sectoral deployments remain with respective ministries while NAICE will maintain interoperability, common standards and monitoring frameworks. See the draft and annexure for proposed council composition: ICT Division (Draft).

Key Focus Areas

The policy identifies multiple focus areas: (1) Smart public services and judiciary—AI to improve service delivery, legal research, case management and access to justice; (2) Data governance, telecommunications and surveillance—standards for interoperable data, computing resources, and regulated surveillance use; (3) Agriculture and food security—precision agriculture, predictive analytics for yields and supply chains; (4) Healthcare—AI-assisted diagnostics, telemedicine and patient monitoring with privacy safeguards; (5) Education, skills and employment—personalized learning and large-scale reskilling programs; (6) Environment, energy and climate resilience—forecasting and resource optimization; (7) Finance and trade—fraud detection, credit scoring with fairness checks; (8) Smart manufacturing and industrial transformation—automation and workplace safety; and (9) Research, innovation and startups—R&D funding, open datasets and incubators. Each sectoral module in the draft outlines illustrative use-cases, likely benefits and principal risks (bias, privacy, safety) and calls for tailored risk mitigation measures and sector-specific roadmaps.

Implementation Framework

The draft recommends a phased implementation model combining enabling measures (standards, shared computing infrastructure, certification), regulatory scaffolding (impact assessments, registration for high-risk systems, procurement rules), and supportive measures (funding, research hubs, curriculum changes). Core procedural steps include mandatory risk assessment for systems proposed by ministries or public bodies, model documentation and provenance (model cards), data management protocols, pilot-to-scale pathways and procurement rules that require verification of vendor compliance. The policy also suggests creation of standardization and certification guidelines for model safety, interoperability and fairness; intellectual property considerations are flagged for follow-on regulation. Implementation will rely on collaboration across the ICT Division, line ministries, NAICE, industry, academia and civil society.

Monitoring and Evaluation

Monitoring will be coordinated by NAICE with performance indicators to measure adoption, safety incidents, inclusivity outcomes, and capacity metrics (trained workforce, research outputs). The draft prescribes periodic reporting, independent audits for high-risk deployments, a national AI registry for documented systems, and a review cycle: the policy will be reviewed at least every three years. The draft also envisages public reporting obligations for significant incidents, a centralized incident reporting channel, and evaluation of social and environmental impacts including energy consumption of large AI models.

Penalties, Liability, and Appeals

As a policy document, the draft defers detailed punitive measures to subsequent legislation (the proposed AI Act) and implementing rules. It recommends a layered liability regime where operators and deployers of high-risk systems carry strict obligations for safety, documentation and remediation; model owners and vendors are subject to contractual and administrative liability; and victims retain civil remedies for harms caused by AI-enabled decisions. The draft also recommends administrative sanctions (fines, suspension of systems, debarment from procurement) for non-compliance and establishes the principle of access to remedies and appeals through administrative channels and courts. The Government's intent to draft a dedicated AI law (with enforcement provisions) has been signalled by the Ministry of Law, Justice and Parliamentary Affairs and reported in public sources: Business Standard / TBS.

Relationship to Other Instruments

The draft is designed to complement existing national instruments such as the ICT Act and Digital Security Guidelines and to align with national strategies (e.g., National Strategy for AI 2019–2024). It explicitly references data governance, cybersecurity frameworks, procurement law and intellectual property regimes, recommending harmonization with a forthcoming Data Protection framework and with sectoral regulations (health, finance, telecom). The policy positions NAICE as the coordinating body to ensure consistency across instruments and to recommend amendments to sector laws where AI-specific gaps are identified.

International Alignment

The policy declares alignment with international norms and ethical frameworks, notably UNESCO’s Recommendation on the Ethics of Artificial Intelligence and UN/UNDP guidance on AI readiness. The draft seeks active engagement with international partners for capacity building, standards harmonization and cross-border incident cooperation. It also calls for benchmarking against global best practices (EU, UK, OECD) and for reciprocal recognition of certifications where appropriate to support trade and technology transfer. UNESCO and UNDP have participated in national readiness assessments that informed the draft (see UNESCO).

Implementation Timeline

Sources and References