Draft AI policy targets innovation, better public services

Bangladesh has drafted a national artificial intelligence (AI) policy to guide the use of AI across all sectors, aiming to become a producer of homegrown innovation, modernise public services, and promote inclusive economic growth.

The draft National AI Policy Bangladesh 2026-2030 provides a formal framework for these goals, aligning the country’s technological ambitions with Vision 2041 and the United Nations’ sustainable development goals (SDGs).

The policy follows an earlier 2024 draft that was not finalised. After the fall of the Awami League-led government and the socio-political changes following the July 2024 uprising, the new draft emphasises digital sovereignty, aiming to safeguard critical data, infrastructure, and citizens’ rights.

DEVELOPING NATIONAL BANGLA LLM

A cornerstone of the draft policy is the development of a Bangla-based advanced national AI system - large language model (LLM) - like OpenAI’s ChatGPT or Google’s Gemini to digitise and preserve Bangladesh’s cultural and linguistic heritage. The model aims to make AI technologies contextually relevant and inclusive, while safeguarding intellectual property from foreign exploitation.

To support such large-scale innovation, the government intends to adopt a “National AI Compute Strategy”, under which centralised graphics processing units (GPUs) will be procured and hosted at the National Data Center for use by various agencies and researchers.

Funding will come from an “AI Innovation Fund”, projected to provide Tk 200 to Tk 250 crore through 2030 for research, development, and commercialisation. Startups and academic institutions will also benefit from targeted tax and customs incentives for importing essential hardware such as servers and accelerators.

RISK-BASED REGULATION OF AI SYSTEMS

The policy introduces a risk-based regulatory framework, categorising AI systems as prohibited, high-risk, limited-risk, or low-risk.

Prohibited applications include social scoring, indiscriminate biometric surveillance, and deepfakes intended to disrupt democracy or elections. High-risk uses, such as in healthcare, law enforcement, and credit assessments, will require Algorithmic Impact Assessments and strict human oversight.

An Independent Oversight Committee, established through an act of Parliament, will audit AI systems for bias and recommend suspension of applications that violate ethical standards or human rights.

The policy also introduces strict liability for high-risk AI, ensuring deployers are accountable for harm regardless of intent.

STEPS TO SAVE JOBS FROM AI THREAT

While AI could increase productivity by 4.3 percent, automation may threaten up to 60.8 percent of garment sector jobs, affecting around 2.7 million workers, and put a total of 5.38 million low-skilled roles across sectors at risk by 2041.

To prepare the workforce, AI education will be introduced from grades 8 and 9, alongside upskilling programmes.

The draft prioritises high-impact sectors such as agriculture and healthcare. AI applications in agriculture will support precision irrigation, pest detection, and localised weather forecasting for Bangladesh’s 16 million farming households.

In healthcare, AI will assist in public health management and crisis prediction, but life-altering clinical decisions will remain under certified medical professionals.

The policy is designed to remain in effect until 2030, when it will be replaced by a permanent Artificial Intelligence Act.

ADDRESSING PRACTICAL PROBLEMS

Faiz Ahmad Taiyeb, special assistant to the chief adviser for the Ministry of Posts, Telecommunications, and Information Technology, said the draft has three primary goals – strengthening AI readiness in institutions, academia, and industry, improving government efficiency through AI, and enhancing service delivery to citizens.

“Additionally, we have tackled risks highlighted in UNESCO’s AI readiness assessment, such as gaps in data protection, interoperability, and cybersecurity. Many of these issues are being addressed through the Cyber Security Act, Data Protection Act, and other initiatives,” he said.

Taiyeb added, “We are improving data access for academia and industry, developing a Bangla LLM to make local knowledge searchable, and creating national interoperability and responsible data exchanges so AI can solve real-life problems effectively.”

He explained that the new policy was needed just a year and a half after the previous draft because the earlier version focused mainly on infrastructure, while the current policy addresses practical problems centred on service delivery.

Md Ashraful Goni, a faculty member at Stony Brook University in New York, praised the draft for positioning Bangladesh as a rights-based, human-centred, and sovereignty-conscious AI nation prioritising ethical governance over rapid commercialisation.

However, he cautioned, “A strong regulatory framework without sufficient technical capacity could unintentionally slow innovation. Risk-based regulation, mandatory algorithmic impact assessments, and centralised oversight could increase compliance burdens for early-stage innovators. AI evolves rapidly, and policy must keep pace.”