The strongest case for Bangladesh's draft amendments is real, and it should be stated first. Online rumors have fueled panic and violence in South Asia, and women and children face harassment, impersonation and AI-generated abuse that existing law handles poorly. Information Minister Andaleeve Rahman has described the amendments as aimed at "disorder on social media" and at "protecting social harmony," according to the Committee to Protect Journalists. A government that inherited a polarized information environment after a political transition is entitled to want tools against it.
The draft, however, does not build those tools. It builds something close to the speech regime Bangladesh has just dismantled.
What the draft would do
At a September 10, 2026 consultation with selected media representatives, the government circulated amendments to the Cyber Protection Act. CPJ reported on September 14 that they would create an offense of circulating online "rumor" or "misinformation" that authorities deem "unsubstantiated or unverified," punishable by up to 10 years in prison or heavy fines. The draft would also broaden takedown powers and let news outlets themselves be convicted. Courts could then suspend or cancel an outlet's operating license, and cases would go to a Cyber Tribunal.
Prothom Alo's reporting on the draft text describes a proposed section 26A. It defines rumor as "unsubstantiated or unverified information" that creates confusion or social instability, and it carries a fine of up to Tk 4 million. That definition is the core problem.
Why 'unverified' is the wrong test
Much of what journalism publishes is unverified at the moment of publication. That is true of a leaked document, an early casualty figure, or an allegation against an official. Verification is a process, and a statute that makes unverified the trigger asks the state to decide when that process is finished. The draft sets no clear standard of proof, intent or demonstrated harm. The person who posts in good faith and is later proven partly wrong faces the same exposure as a deliberate fabricator.
International law asks for more. Article 19 of the Universal Declaration of Human Rights protects the right to "seek, receive and impart information and ideas through any media." The UN Human Rights Committee's General Comment 34 is the authoritative reading of the equivalent ICCPR guarantee. Under it, restrictions must be provided by law, serve a legitimate aim and be necessary and proportionate. A vague, open-ended offense enforced by the same authorities whose conduct is being reported falls short of that test.
The license-cancellation power raises the stakes further. A 10-year sentence deters individuals. Letting courts close down an outlet deters every newsroom at once, and it does so without anyone having to prove that a specific claim caused specific harm.
The record the draft is walking back
Bangladesh has run this experiment. CPJ counts at least 255 journalists prosecuted under the 2018 Digital Security Act. The writer Mushtaq Ahmed died in custody after nine months of pretrial detention under it.
The post-transition reforms were a real achievement. The Cyber Security Ordinance 2025, gazetted on May 21, 2025, omitted nine sections of the 2023 Act: 21, 24 through 29, 31 and 34. It annulled pending cases under them and recognized internet access as a civic right. Those sections had covered offenses such as distorting the liberation-war narrative, offensive communications and defamation. On October 9, 2025, the Advisory Council approved amendments that allow all Digital Security Act cases to be dismissed. CPJ now warns that the draft revives Digital Security Act-style speech offenses. The record supports it: the draft brings back a broad, discretionary speech crime after the state had decided to repeal one.
Better instruments exist
Proportionate regulation is possible, and it does not start with a general ban on "unverified" content. A workable approach would have four parts.
- Target harm, not uncertainty. Criminalize knowing falsehoods that cause specific, demonstrable harm, such as incitement to violence or fraud. Require proof of intent, and leave good-faith error and opinion outside the offense.
- Protect journalism explicitly. Exempt public-interest reporting and set a standard of reasonable diligence. Civil corrections and right-of-reply remedies do the job without a prison sentence.
- Put courts before takedowns. Takedown orders should be reviewable and time-limited, and they should come with notice to the publisher.
- Regulate the real online-safety harms directly. Non-consensual imagery, child exploitation and impersonation are already addressed in the 2025 ordinance. Enforcing those provisions well would do more for safety than a new rumor offense.
There is also a wider pattern of weakening safeguards. A July 28, 2026 joint statement from Access Now, Amnesty International, ARTICLE 19, Human Rights Watch and others credited the reform agenda. It also warned that the Bangladesh Telecommunication Regulation (Amendment) Act, 2026 removed safeguards. It put spending on interception and monitoring systems at roughly $190 million between 2015 and 2025. Pairing expanded surveillance capacity with a vague speech offense is the combination that produced the abuses the reforms were meant to end.
The innovation cost
Vague liability chills more than journalism. Platforms, startups and creators operating in Bangladesh cannot price the risk of a law whose trigger is an official's view of what is verified. The predictable responses are over-removal, self-censorship and a retreat from the market. Those outcomes hurt a digital economy the country is trying to grow.
The consultation gives the government a chance to change course. It should withdraw section 26A or rewrite it around intent and demonstrated harm. It should drop the license-cancellation power and publish the full text for open comment. The 2025 reforms showed that Bangladesh can repeal a bad speech law, and keeping that repeal in place is the stronger way to protect its online public.