Bangladesh's First Responsible AI Summit Kicks Off at MIST in Dhaka

MIST hosted Bangladesh's first Responsible AI Summit 2026, focusing on local context, policy, and workable systems. Speakers and students set next steps; top teams earned awards.

MIST hosts Bangladesh's first Responsible AI Summit 2026

The Military Institute of Science and Technology (MIST) convened Bangladesh's first Responsible AI Summit at Dhaka Cantonment, led by the Department of Computer Science and Engineering. The event set a clear agenda: build AI that fits local realities, meets public expectations, and supports policy with evidence.

Women and Children Affairs Adviser Sharmin S. Murshid, also in charge of the Ministry of Social Welfare, opened the summit as chief guest. MIST Commandant Major General Mohammad Nasim Parvez served as Chief Patron, with Brigadier General Mohammad Shahjahan Mojib, Head of CSE, presiding over the sessions.

Key voices and themes

Sharmin S. Murshid underscored the role of responsible AI in industrial growth and practical policy. Major General Parvez affirmed MIST's commitment to push education and research that translate into deployable, accountable systems.

Dr. Upol Ehsan (Northeastern University), also a Berkman Klein Fellow at Harvard and a known name in Human-Centered Explainable AI (HCXAI), delivered the keynote. Contributors in subsequent sessions included Associate Prof Dr. SM Tayyabul Haque (BRAC University), Prof Dr. Nova Ahmed (North South University), Associate Prof Dr. Kazi Noor-e-Alam Siddiqui (MIST), Managing Director and Co-Founder of Bondstein Technologies Mir Shahrukh Islam, CEO and Chairperson of UY Systems Limited Farhana Anwar Rahman, Senior Assistant Secretary of the ICT Department Arafat Mohammad Noman, and Managing Director of Graphic People and Software People Imtiaz Elahi.

Why it matters for builders and researchers

Local context first: models must handle Bangla, regional dialects, and sector-specific data common in Bangladesh.
Risk and policy: teams need clear audit trails, human oversight, and deployment controls that map to public-sector and enterprise requirements.
Resource constraints: efficient inference, offline modes, and cost-aware pipelines are as important as accuracy.
Trust by design: user-facing transparency (model cards, data-use notes, known limitations) reduces friction in adoption.

Discussion highlights

Build indigenously where it counts-especially for language, data access, and infrastructure constraints unique to Bangladesh.
Academia's role spans independent evaluation, dataset stewardship, and educating a workforce that can ship safely.
Responsible use in a developing-country context requires practical guardrails: bias checks on local data, stress tests for edge cases, clear escalation paths for incidents.

Student competition and awards

Students from 12 universities took part, including MIST, Dhaka University (DU), Bangladesh University of Engineering and Technology (BUET), BRAC University, Bangladesh University of Professionals (BUP), and Islamic University of Technology (IUT). Awards were presented to the top four teams in project presentations and the top four teams in poster exhibitions.

What teams can do next

Adopt a lightweight Responsible AI policy and risk register; review at each model update.
Use model cards and data sheets; document known failure modes and fallback steps.
Evaluate with local datasets; measure fairness across language, region, and access patterns.
Set up a cross-functional review (engineering, legal, operations) for pre-deployment checks.
Define an incident response playbook for model errors, escalation, and user communication.
For public-facing systems, add human-in-the-loop for high-impact decisions.