Bangladesh's Digital Sovereignty Is About the Right to Leave a Vendor

Bangladesh has tied more of daily life to digital platforms: mobile financial services handle small transactions, online marketplaces connect local sellers to city buyers, and freelancing platforms support many young earners. Government services, banking, education and commerce now lean heavily on digital infrastructure. According to Daily Star, the pressing question is how much control the country keeps over the systems its economy runs on.

The opinion piece argues that control is not only about data protection compliance or where servers are physically located. It is about structural ownership: who owns the core code, who sets the rules, who captures the economic value, and whether Bangladesh can choose a different path when current vendors no longer serve national interests. That ability to choose is described as the real test of digital sovereignty.

Servers in Dhaka are not the same as ownership

The article says public debate has focused narrowly on server location and data localisation mandates. Keeping sensitive information close is a valid regulatory response, but locally hosted cloud platforms can still run on proprietary foreign software, licences and support, leaving core architectural control outside domestic oversight. At the same time, building every technical asset at home is described as neither viable nor economically sensible for a developing market.

The regulatory picture changed, the piece notes, when the Personal Data Protection (Amendment) Ordinance was signed in February 2026, narrowing the in-country data copy requirement to restricted data and critical information infrastructure. Days later, Bangladesh signed a trade agreement with the United States committing to free data transfer across trusted borders and to avoiding policies that discriminate against US digital products. With location no longer a reliable policy lever, the argument turns to provider mobility: for an emerging economy, sovereignty is measured by the cost and speed of leaving a vendor. The capacity to exit creates bargaining leverage, while lock-in reduces a country to a data tenant wherever its servers sit.

Even a sovereign government cloud platform, with dedicated-region hardware and greater local control, must answer a basic question: if the partnership sours or priorities diverge, how smoothly can workloads move elsewhere, and has that migration ever been tested?

Mobile number portability shows the gap

The article offers the case of a young woman running a boutique fashion business from a rented flat in Mirpur. Clients order by phone and payment confirmations arrive by text message. Seeking better coverage, she moved the business number to a new operator through mobile number portability. Call quality improved at once, but the one-time passwords needed to verify payments stopped arriving, and within days she was back in a customer care queue asking to reverse the switch.

Her experience reflects a wider pattern, according to the piece. Mobile number portability promised easy switching, yet figures presented at a telecommunications seminar in April 2025 showed that less than 1% of active subscribers had used the service in over six years, and that 45% of switching requests were rejected by the original operators. Because text messages lacked a mandatory interconnection layer and mobile financial service providers were not required to route one-time passwords through the central platform, ported numbers routinely missed critical payment alerts. After the first year, more than half of those who tried to switch had already failed. A legal right to exit without supporting technical architecture amounts to systemic lock-in.

The revenue system that had no exit plan

The risk reaches core state operations, the article says. In 2015 the National Board of Revenue signed a World Bank-funded $33.6 million deal with a Vietnamese firm to automate value-added tax administration, the government's largest source of domestic revenue. Designed for a one-year development phase and a five-year maintenance period, the project instead ran into repeated contract extensions. Two later tenders to hand operations to a local firm drew no qualified bidders, and an NBR official acknowledged that the original agreement had no provision for transferring the system to local teams. The country built a core revenue system without an exit strategy, which the piece uses to argue for smarter procurement that builds domestic ownership rather than permanent dependency.

What other markets have tried

International experience, according to the article, shows that preventing vendor lock-in needs regulatory oversight. Through its Data Act, the European Union made cloud switching a legal right, with switching fees set to be banned outright from January 2027. Europe also funds domestic cloud alternatives, but the law's strength is described as making migration cheaper and smoother so competition protects user autonomy. Such rules must still be balanced to avoid market distortions or trade friction.

In India, even the successful UPI payment network struggles with concentration, with two apps controlling over 80% of transactions while regulators repeatedly deferred the 30% market-share cap. Brazil illustrates the trade risk, having faced 25% US tariffs in July 2026 after an investigation targeted advantages given to its central bank's Pix system. Policies favouring national champions can draw retaliation, while rules that make every provider equally easy to leave are harder to label discriminatory.

The article calls for strict portability and interoperability mandates in the operating procedures of the Personal Data Protection Act, with clear migration deadlines, capped transfer fees, penalties for unjustified rejections, and a requirement that banks and wallets route one-time passwords to a user's new carrier. It also stresses, as critics noted when the amended ordinance was gazetted, that true sovereignty requires privacy rules binding state agencies as strictly as private companies.

The sovereign AI question

The tension is sharpest in the state's emerging sovereign AI plans, the piece argues. After the government announced in October that sovereign AI infrastructure would launch within a year, reports emerged that Dhaka is seeking a $250 million Asian Development Bank loan, possibly co-financed by Korea, to fund a dedicated Bangla large language model using LG's EXAONE model, FuriosaAI chips and NAVER cloud infrastructure. No agreement has been signed. Relying on one foreign ecosystem for models, chips and cloud services risks repeating the VAT automation trap on a far greater scale, especially since existing open-weight models can already process Bangla. The real question, the article says, is not whether to build but whether the country can leave.

True computational sovereignty, in this view, does not mean training a model in an isolated silo so that it speaks Bangla. It means owning the underlying infrastructure and data pipelines while keeping the freedom to swap the models and chips that run on them. If the state finances specialised hardware and data centres without a clear data strategy and exit plan, it risks another high-cost turnkey system dependent on external updates. The policy objective should shift from building everything at home towards infrastructure, data control and private-sector enablement, including procurement that requires a verified exit test before final sign-off, with vendors demonstrating that production workloads, databases and platform assets can be migrated to an in-house team or an alternative provider.

Sources: The Daily Star