Bangladesh closed 2025 with a regulatory paradox: the same year the government took meaningful steps to reform a notorious cyber-speech law, it also pushed through legislation granting regulators sweeping, poorly-checked powers over every major online platform operating in the country. The two moves are not in balance — and the architecture of the second one should concern anyone who watched what the first law's predecessor became.
From DSA to CSA to CSO: A Statutory Trail
To understand where Bangladesh stands, the legislative genealogy matters. The Digital Security Act 2018 became one of the region's most widely condemned content-moderation statutes. Its vague provisions on defamation, criticism of the Liberation War, and content deemed offensive to religious sentiments were routinely used to prosecute journalists, activists, and ordinary social media users. By the time the interim government took power in 2024, thousands of cases had been filed under it.
Parliament replaced the DSA with the Cyber Security Act (CSA) in September 2023 — but civil society swiftly noted the CSA retained the DSA's most abusive provisions under new branding. The genuine break came in May 2025. On May 6, Law Adviser Asif Nazrul announced the repeal of nine of the CSA's most controversial sections, including those criminalising commentary on the Liberation War, defamation of Sheikh Mujibur Rahman, and insults to national symbols. The Cyber Security Ordinance 2025 — officially Ordinance No. 25 of 2025 — was gazetted on May 21. All remaining speech offences became bailable, with a maximum two-year sentence. In October 2025, a further amendment formally cancelled all outstanding DSA-era cases, ending proceedings that had hung over hundreds of defendants for years.
These are real reforms. They were won through sustained civil society pressure and international scrutiny, and they mark a meaningful departure from the DSA's worst legacy.
The Telecom Amendment: A Different Architecture
The Bangladesh Telecommunication (Amendment) Ordinance 2025, approved December 24, tells a more complicated story. The Bangladesh Telecommunication Regulatory Commission's mandate was dramatically expanded to cover social media platforms, OTT services, streaming video, television, AI services, cloud providers, data centres, e-commerce, and online payment systems — sectors previously operating outside its formal reach.
Platforms classified as intermediaries must now fulfill content removal requests from government agencies within 72 hours. Failure carries penalties of up to Tk 300 crore (approximately USD 35 million) and up to five years' imprisonment — penalties that can be imposed repeatedly for ongoing violations. ARTICLE 19, which analysed the ordinance in detail, warned that the definitions of "intermediaries" are broad and vague enough to enable sweeping content removals without adequate judicial authorisation. The group identified "minimal safeguards and no requirement for judicial authorization" across large swaths of the enforcement mechanism.
One genuine win in the final text: a permanent ban on internet and telecom shutdowns under any circumstances — a hard-fought provision that reflects lessons from costly shutdowns during earlier political unrest.
The Context That Complicates Easy Dismissal
Steelmanning the government's position requires acknowledging a documented, serious problem. On December 18–19, 2025, mobs attacked and set fire to the offices of the Daily Star and Prothom Alo — Bangladesh's two most prominent newspapers. Over 26 staff members were temporarily trapped before being safely evacuated. Investigations by both outlets and the fact-checking organisation Dismislab found that calls for violence had circulated openly on Facebook for months before the attacks, including through a group with 68,000 members. The first widely-shared call to action appeared on December 15 — three days before the attacks.
The BTRC formally wrote to Meta demanding urgent action and Bengali-language moderation capacity — a reasonable request that exposed the real structural gap between global platform policy and local-language harm. In March 2026, Amnesty International documented the same dynamic ahead of parliamentary elections: cross-border inflammatory content from India targeting minorities and political parties, algorithmically amplified at scale, with Meta's Bangla-language response capacity badly lagging. The OHCHR weighed in following the December attacks, calling for urgent accountability and noting the violence constituted a direct assault on press freedom.
These are not hypothetical harms. The connection between inadequate platform moderation and offline violence in Bangladesh is documented and severe. Any responsible regulatory response has to engage with this reality.
Where Proportionality Breaks Down
The problem is not that Bangladesh wants platforms to act more swiftly on incitement to violence. The problem is that the Telecom Amendment's enforcement design — vague definitions, 72-hour removal windows, USD 35 million penalties, limited judicial review — creates infrastructure easily weaponised for purposes well beyond genuine safety.
The DSA's history makes this risk concrete rather than abstract. Laws drafted in the language of security and public order have repeatedly been turned against the press, political opponents, and religious minorities. Rebuilding trust after that record requires not just better-worded statutes but structural constraints on executive discretion — mandatory judicial authorisation before takedown orders take effect, transparent appeals mechanisms, and statutory definitions narrow enough that a regulator cannot quietly expand scope over time.
Meta's failures in Bangladesh are real and require sustained pressure. But the mechanism for applying that pressure matters enormously. Giving an authority with BTRC's institutional history unlimited discretion over the full stack of the internet economy — social media, AI, cloud, payments — without meaningful judicial checks is not a solution to platform impunity. It creates a parallel instrument of control operating under a different name.
What Proportionate Accountability Looks Like
A framework that actually serves Bangladeshi users would look different: a statutory definition of harmful content limited to categories with direct, documented links to physical harm (incitement to violence, coordinated targeting of individuals); mandatory court authorisation before any removal order reaches a platform; a functioning appeals process accessible to both platforms and affected users; and direct investment in Bengali-language trust and safety capacity — either through bilateral agreements with major platforms or through funded local civil society infrastructure — rather than fines that fall heaviest on smaller services.
Bangladesh demonstrated in 2025 that its government can move quickly when accountability demands it. Nine abusive provisions were stripped from the statute books, hundreds of cases dismissed, a shutdown ban enacted. The open question for 2026 is whether that same energy applies to making the Telecom Amendment's platform powers proportionate. As currently structured, the ordinance is a capable instrument of censorship wearing the vocabulary of platform safety. Getting to a durable framework means narrowing scope, mandating judicial oversight, and investing in regulator competence — not just regulatory reach.