What the Parliament reply actually says
In a written reply reported on 4 November 2026, Communications Minister Fahmi Fadzil told the Dewan Rakyat that more than 117,000 harmful posts were removed after requests from the Malaysian Communications and Multimedia Commission (MCMC) between 1 January 2022 and 15 September 2026. According to The Star's report of the reply, the total breaks down into 60,632 bullying-related items, 55,895 false-content items and 1,277 defamatory posts. Those three figures sum to 117,804. The reply cites the Online Safety Act 2025 (Act 866) and notes that defamation and cyberbullying offences are enforced by police under the Penal Code. The minister framed the approach as consistent with constitutional free expression.
A detail matters here. The Star reports that the bullying items were taken down by MCMC, while the false-content and defamation items were removed by service providers after MCMC requests. The Sun's account attributes all removals to MCMC requests. The reply is not a clean ledger of who decided what, which is itself a transparency gap.
The strongest case for the government
The case for acting is real. Cyberbullying can drive young people to self-harm, and a victim of a viral false claim cannot wait for a civil suit to run. A regulator that asks platforms to remove content, rather than prosecuting speakers, is arguably a lighter touch than the criminal route. Reporting the numbers to Parliament at all is also more transparency than many governments offer. The Online Safety Act 2025 came into force on 1 January 2026, and MCMC's Online Safety Act page lists the harms it targets: child sexual abuse material, financial scams, violent or terrorist content, self-harm inducement and harassment, among others. Much of that list is uncontroversial, and a pro-speech publication should say so.
Why the headline number proves little
The difficulty is that 117,804 is a throughput figure. It tells readers how much was removed, not whether it should have been. Three things are absent from the reply as reported.
First, there is no error rate. We do not know how many of the removed items were later restored, appealed, or found lawful. Takedown counts rise when thresholds are loose and when requests are sent in bulk. A rising count is therefore compatible with better enforcement, worse enforcement, or just more requests.
Second, the categories are uneven in how objectively they can be judged. Child abuse material is binary. "False content" and "defamatory" are contested judgments that courts normally make after hearing both sides. Here 55,895 false-content items were removed through administrative requests. Defamation in particular is a civil and criminal matter in which the speaker usually has a chance to respond. A takedown request to a platform skips that step.
Third, the volume itself is a warning about process. In July 2026 Fahmi told Parliament that MCMC had identified 345,712 pieces of harmful content for removal between 1 January and 1 July 2026, and that each request takes 30 to 45 minutes of staff time, equal to about 19.7 years of working time at the lower figure. The same report says roughly 91% of that flagged content was gambling and scam-related. That is a different measure from the 117,000 total and should not be added to it. But it shows what the pipeline is mostly handling: fraud, not opinion. A regime that is mainly fighting scams should say so, and should keep its speech-adjacent categories (false content, defamation, bullying) under tighter, separately reported procedures.
The legal architecture that sits behind the numbers
Two statutes carry the weight. The Communications and Multimedia (Amendment) Act 2025 took effect on 11 February 2025. According to Azmi & Associates' overview, it changed section 233 of the 1998 Act by replacing "offensive" with "grossly offensive" and raised the maximum fine from RM50,000 to RM500,000. Narrowing the adjective is a genuine improvement over the older, vaguer wording. A tenfold rise in the fine cuts the other way.
The Online Safety Act 2025 applies to licensed service providers rather than individuals. Per Skrine's January 2026 summary of the MCMC FAQ, platforms with eight million or more Malaysian users are deemed licensed, and a person aggrieved by an MCMC instruction or decision may appeal to an Online Safety Appeal Tribunal. That appeal route is the right design instinct. But the minister's reply reports no tribunal statistics, and the Act's regulations and text are published on MCMC's legislation page without any accompanying enforcement dataset.
What proportionate oversight would look like
The Electronic Frontier Foundation's recent argument about election-period moderation applies well beyond elections: perfect content moderation at scale is impossible, so the question is whether a system is transparent, appealable and honest about errors. Malaysia's own framework has the pieces. It needs to publish them.
A practical reporting standard would be short. MCMC should publish, at least semi-annually, the number of requests by category and platform; the share complied with; the number restored after appeal or review; and tribunal outcomes. It should separate scam and child-safety removals from false-content, defamation and bullying removals, since the speech risk differs sharply. And it should state whether a request was a legal order or an informal ask, because platforms treat the two differently.
None of this slows enforcement against fraud or child abuse. It does make it harder for a large number to stand in for a justified one. Fahmi is right that Parliament should hear the figures. The next written reply should include the denominator: how many removals were wrong, and who found out.