Bangladesh National AI Strategy — 12 Pillars for 2026–2033

Publication Date: January 2026

Classification: Strategy Document

Scope: National-level; 7-year horizon (2026–2033)

---

Preamble

Bangladesh faces a defining decision. The global AI transition is accelerating and the choices made in the next 36 months will determine whether Bangladesh participates in AI as a creator, as a user, or as a dependent consumer of technology made elsewhere for priorities that are not Bangladesh's own.

This strategy document proposes a 12-pillar national AI framework designed for Bangladesh's specific capabilities, constraints, and aspirations. It is grounded in evidence, costed against Bangladesh's fiscal realities, and sequenced for implementation within existing institutional structures.

The strategic objective: Raise Bangladesh's AI Readiness Score from 47.12 (2024) to 70+ by 2033, and position Bangladesh as the leading AI hub in South Asia outside India by 2035.

---

Strategic Context

Why 70+ Matters

The Oxford Insights AI Readiness Index 70+ threshold separates nations that are doing AI from nations that are becoming AI-native. At 70+, countries typically have:

Functional national AI regulatory architecture
Self-sustaining domestic AI research community
Multiple globally competitive AI-first companies
Government services routinely augmented by AI
AI literacy as a standard component of education

Vietnam (58.21), Rwanda (52.16), and Malaysia (72.4) illustrate different pathways. Malaysia's 72.4 score reflects 15 years of sustained investment. Bangladesh can reach 70+ in 7 years through focused, sequenced investment.

What Must Change

Bangladesh's critical weakness is its Technology Pillar score: 26.26/100. This reflects:

Near-zero domestic AI research output (publications, patents)
No sovereign compute infrastructure
Bangla language severely underrepresented in global AI training data
AI talent pool concentrated in Dhaka with significant brain drain to North America and the Gulf

Government (58.52) and Data (56.59) scores are moderate — the foundation exists for rapid improvement if the technology gap is addressed.

---

The 12 Pillars

Pillar 1: National AI Commission (NAC)

What: A statutory body with constitutional mandate, cross-ministerial authority, and a dedicated budget line — not a coordination committee, but an executing agency.

Why: Bangladesh has had multiple "national AI initiatives" that failed because they had no authority, no budget, and no accountability. The NAC must have real power.

Structure:

Chair: Cabinet Minister (reporting directly to Prime Minister)
Vice Chairs: Finance Secretary, ICT Division Secretary, BUET Vice Chancellor
Executive Director: Full-time professional (competitive international appointment)
Budget: Tk 500 crore annually (Year 1), scaling to Tk 2,000 crore by Year 5

Mandate:

Coordinate AI implementation across all 20 ministries
Approve and monitor National AI Strategy implementation
Report to Parliament annually
Authority to direct ministry-level AI investment allocations

Timeline: Establish by legislation within 12 months of strategy adoption.

---

Pillar 2: Sovereign AI Compute Infrastructure

What: A national GPU cluster — Bangladesh's own AI compute capacity, not dependent on foreign cloud providers.

Why: AI training requires massive compute. Every taka spent on foreign cloud is a taka that builds no domestic capability. Countries that control compute control AI.

Target: 1,000 NVIDIA H100-equivalent GPUs (or equivalent) by 2028; 5,000 by 2031.

Implementation:

Phase 1 (2026–2027): 200-GPU cluster at BUET/DU AI Research Institute — Tk 300 crore
Phase 2 (2027–2029): Scale to 1,000 GPUs through National AI Research Cloud — Tk 800 crore
Phase 3 (2029–2031): Full national AI cloud for government and research — Tk 2,000 crore

Operational model: Hybrid — government core (for defence, public sector, classified research) + commercial tier (universities, startups, companies at subsidised rates).

Funding options: 40% ADP allocation; 40% multilateral development bank loans (ADB, World Bank AI for Development); 20% public-private partnership with Bangladesh's own tech companies.

---

Pillar 3: Bangladesh AI Research Institute (BAIRI)

What: A world-class AI research institution — modelled on the Alan Turing Institute (UK) or Mila (Canada) — permanently funded, internationally recruited, publishing in top-tier venues.

Why: Bangladesh currently produces fewer than 15 peer-reviewed AI publications per year in top venues. BAIRI changes this.

Structure:

Hosted jointly at BUET and Dhaka University
30 core researchers (10 international recruits, 20 Bangladesh nationals including diaspora returns)
200 PhD fellowships funded (2026–2033)
Annual budget: Tk 200 crore, growing to Tk 500 crore
Mandate: publish internationally AND solve Bangladesh-specific problems

Priority research domains:

Bangla NLP and speech recognition
Climate AI and disaster resilience
Agricultural AI for smallholder farmers
Health AI for resource-limited settings
AI governance and safety

Revenue model: 30% government grant; 30% international research grants; 40% contract research for government ministries and private sector.

---

Pillar 4: Bangla AI Sovereign Stack

What: A suite of foundational Bangla-language AI tools — open-source, freely available to all — that make Bangladesh's language a first-class AI language.

Components:

BanglaBERT 2.0: Updated and extended language model with 10× more training data
BanglaSTT: Speech-to-text with 95%+ accuracy for all Bangladeshi dialect variants
BanglaTTS: Text-to-speech for citizen services and accessibility
BanglaOCR: Handwritten and printed Bangla document recognition at 98%+ accuracy
BanglaGPT: Instruction-following language model for government and education use
Bangla Multimodal: Image + text understanding for rural citizen services

Why this matters: If Bangladesh does not build these tools, foreign companies will — and they will be trained on data that does not represent Bangladesh's dialects, contexts, or needs.

Governance: Open-source under CC0/Apache 2.0 license; maintained by BAIRI; accessible to any developer in Bangladesh.

Timeline: BanglaBERT 2.0 and BanglaOCR by 2027; BanglaSTT and BanglaTTS by 2028; BanglaGPT by 2029.

---

Pillar 5: National AI Education Strategy

What: AI literacy embedded in Bangladesh's education system from primary school through university, with a focus on creating AI practitioners, not just AI users.

Components by level:

Primary (Class 1–5): Computational thinking, logic games, basic data concepts. No coding required. Target: 10M students by 2028.

Secondary (Class 6–10): Introduction to algorithms, basic Python, AI concepts and ethics. Target: 8M students by 2029.

Higher Secondary (Class 11–12): AI applications, data science basics, Bangla NLP projects. Target: 3M students by 2030.

University undergraduate: AI/ML courses mandatory for all engineering and science programs; elective for social sciences and arts. Target: 500,000 students per year by 2028.

Graduate: 5,000 AI/ML master's and PhD seats annually by 2030 (current: ~200).

Technical and vocational: 100,000 AI-adjacent technical workers trained annually (data labelling, AI system maintenance, model fine-tuning) — this is Bangladesh's near-term employment opportunity.

Teacher training: 50,000 teachers trained in AI basics by 2028 through DPEd and BED curriculum updates.

---

Pillar 6: National AI Data Infrastructure

What: The data foundations that every other pillar depends on — a national data strategy, data quality standards, and shared data infrastructure.

Key components:

National Data Exchange (NDEx): A secure government API platform enabling inter-ministry data sharing with role-based access control. Priority integrations: NBR + Bangladesh Bank + BRTA + Election Commission + MoLGRD for citizen identity verification.

National AI Training Data Repository: 100TB+ of high-quality Bangla-language data for AI training — government documents, court records, literature, news (with rights clearance). Freely available to BAIRI and licensed researchers.

Data Quality Standards: NAISB certification for datasets used in government AI training. Minimum standards for representation, accuracy, and recency.

Open Data Portal v2.0: data.gov.bd upgraded to provide machine-readable, API-accessible data across all major government datasets by 2027.

---

Pillar 7: AI Startup and Innovation Ecosystem

What: Bangladesh has 4,000+ tech startups; fewer than 50 are AI-first. This pillar changes that.

Components:

Bangladesh AI Fund: Tk 500 crore venture-style fund investing in AI startups at seed stage (Tk 50L–2Cr) with no equity requirement for early stage, equity at growth stage. Focus: agriculture, health, finance, climate, education AI.

AI Innovation Zones: Special economic zones in Dhaka, Chittagong, and Sylhet with subsidised compute access, talent visa fast-track for foreign AI experts, and regulatory sandbox access.

AI in Procurement: Bangladesh government must allocate 15% of ICT procurement budget to AI products and services from locally incorporated companies by 2028.

Global diaspora program: 500 Bangladeshi AI professionals returning from abroad over 7 years with salary bridging support for the first 2 years.

---

Pillar 8: AI in Government Services (GovAI)

What: AI deployed across government services to improve efficiency, reduce corruption, and improve citizen experience.

Priority deployments (2026–2027):

1. NBR tax compliance scoring

2. MoHFW disease surveillance AI

3. Land registration fraud detection

4. Flood early warning system upgrade

5. MoA crop monitoring and advisory

Standards: All GovAI deployments must comply with Bangladesh National AI Standards (BNAS), include human review mechanisms for decisions affecting individual rights, and publish annual performance reports.

Target: 50 government AI deployments operational by 2028, covering 20M+ citizen interactions annually.

---

Pillar 9: AI in Healthcare

What: A dedicated healthcare AI strategy within the national AI framework.

Priority use cases:

Telemedicine AI triage for community clinics
Radiology AI assistance (TB, cancer screening)
Maternal mortality risk prediction
Drug supply chain optimisation
Medical record digitisation and standardisation

Why separate: Health AI has specific ethical and safety considerations requiring dedicated governance — the DGDA needs AI-specific medical device approval pathways.

Target: AI assists in clinical decisions for 10M outpatient interactions annually by 2030.

---

Pillar 10: AI Safety and Ethics Framework

What: Bangladesh's specific approach to ensuring AI is safe, fair, and aligned with Bangladeshi values.

Bangladesh AI Ethics Charter (proposed):

1. AI shall not discriminate on the basis of religion, gender, district of origin, or economic status

2. AI decisions affecting citizens' rights must be explainable and contestable

3. Bangladesh's digital sovereignty — control over data about Bangladeshis — is non-negotiable

4. AI must serve the 170 million, not just the urban 10 million

5. Bangladesh will not deploy AI for autonomous lethal weapons

Enforcement: Bangladesh National AI Authority (BNAIA) — see Pillar 11.

---

Pillar 11: Regulatory Architecture

What: The legal and institutional framework for AI governance.

Key institutions:

Bangladesh National AI Authority (BNAIA) — independent regulator
National AI Standards Board (NAISB) — technical standards
Parliamentary AI Oversight Committee — legislative accountability

Key legislation:

National AI Act (target: 2027)
AI-specific implementing rules under Personal Data Protection Act (target: 2026)
Government AI Procurement Rules (target: 2026)

(See separate policy paper: "AI Policy & Regulatory Framework for Bangladesh")

---

Pillar 12: International AI Partnerships

What: Strategic AI relationships that accelerate Bangladesh's progress without creating dependency.

Priority partnerships:

Bilateral:

India: Joint Bangla NLP research (shared language); AI in disaster management
Singapore: Financial AI, regulatory sandbox experience-sharing
South Korea: Manufacturing and industrial AI
UAE: Smart city AI, fintech
UK: FCDO AI development program; Turing Institute collaboration

Multilateral:

UNESCO: AI Ethics implementation
ITU: AI for Development program
World Bank: AI skills development financing
ADB: Digital infrastructure investment
G77: Developing-nation AI governance coalition

Diaspora science network: Formal network of Bangladeshi AI researchers abroad (estimated 2,000+ globally) with annual summit and collaboration grants.

---

Funding Overview (7-year total: Tk 8,500 crore)

| Pillar | 7-Year Budget (Tk crore) | Primary Source |

|--------|--------------------------|----------------|

| 1. National AI Commission | 200 | Government revenue |

| 2. Sovereign Compute | 3,000 | ADP + MDB loans |

| 3. AI Research Institute | 800 | Government + grants |

| 4. Bangla AI Stack | 300 | Government + open source |

| 5. AI Education | 1,000 | Education budget + MDB |

| 6. Data Infrastructure | 400 | Government ICT budget |

| 7. Innovation Ecosystem | 700 | Public-private fund |

| 8. GovAI Deployments | 600 | Ministry budgets |

| 9. Health AI | 400 | MoHFW + WHO |

| 10. Safety & Ethics | 100 | Government |

| 11. Regulatory Architecture | 300 | Government |

| 12. International Partnerships | 200 | Government + aid |

| Total | 8,000 | |

Tk 8,000 crore over 7 years = approximately $720M USD. This is 0.15% of projected 7-year GDP — comparable to what Singapore, Vietnam, and Kenya have committed to national AI strategies.

---

Accountability Framework

Annual Progress Report: National AI Commission publishes progress against all 12 pillars each year. Tabled in Parliament.

3-Year Review: Independent international review of strategy progress. Findings made public.

Key Performance Indicators (2033 targets):

Oxford Insights AI Readiness Score: 70+ (from 47.12)
Annual AI publications in top venues: 200+ (from <15)
AI startups (active, revenue-generating): 500+ (from ~50)
Government AI deployments: 100+ (from 0)
Annual AI graduates: 10,000+ (from ~200)
Bangla NLP accuracy benchmark: 95%+ (from ~70%)
ICT exports including AI: $5B+ (from $1.4B)

---

Conclusion

A 70+ AI readiness score by 2033 is ambitious but achievable. Vietnam moved from 42 to 58 in 5 years with disciplined investment. Rwanda moved from 38 to 52 in 6 years with strong political will. Bangladesh has advantages both countries lacked: a larger domestic market, a significant diaspora, an existing ICT industry, and the urgent motivation of a 170-million-person nation that cannot afford to be left behind.

What Bangladesh lacks is coordinated will. This strategy provides the architecture. The rest depends on political commitment to execute it.

Strategy document updated January 2026. Comments and submissions welcome at research@bangladeshai.org