Malaysia's Communications and Multimedia Commission (MCMC) issued more than a million content takedown requests to online platforms between January 1, 2025 and August 31, 2026. Platforms complied with 935,416 of them — a 93% compliance rate that, on paper, looks like a regulatory success story. Communications Minister Fahmi Fadzil chose instead to highlight the other number: 65,280 pieces of scam and illegal gambling content that stayed online anyway.
Fahmi delivered the figures in his opening address at the inaugural M360 ASEAN 2026 conference in Kuala Lumpur (September 9–10). Of the content actually removed, 592,695 items were gambling-related and 227,486 were scam-related — together about 88% of all takedowns. His message to platforms was blunt: "Malaysia's position is firm. We will not budge. They need to be more proactive, not merely remove [content]." He pointed separately to sellers using coded language to evade detection on fentanyl-laced vape products, and to earlier MCMC data showing Facebook alone accounted for 78% of detected gambling content and 53% of scam content on Malaysian platforms.
The Case for Pushing Platforms Further
The minister's frustration deserves to be taken seriously before it's argued against. A 93% compliance rate sounds high until you remember it was measured against volume: 65,280 outstanding pieces of content is not a rounding error, it's a functioning distribution channel for financial fraud and unlicensed gambling that keeps operating in the gap between a complaint and a removal. Scam networks don't need every post to survive — they need enough of them to survive long enough to extract money from victims before enforcement catches up. A notice-and-takedown regime, by construction, always leaves a residual layer of harm sitting in that lag. Regulators who watch that residue compound at scale, month after month, have a legitimate case that reactive processing alone cannot solve a problem this size, and that platforms with far more granular access to their own detection signals are better positioned to catch coded evasion — like disguised vape-and-fentanyl listings — before MCMC ever has to ask.
Why "More Proactive" Is a Harder Ask Than It Sounds
But the leap from "platforms should catch more scams" to "platforms should proactively filter more content" runs into a problem specific to Malaysia's legal architecture: the standard they'd be filtering against is not well-defined. Section 233 of the Communications and Multimedia Act 1998 — the underlying statute MCMC's takedown authority traces to — criminalizes content that is "obscene, indecent, false, menacing or offensive in character," language civil society groups including Sinar Project and the Malaysian Bar have long criticized as vague enough to have been used against a satirical musician, a whistleblower news site, and ordinary government critics, not just scam networks. Asking platforms to move upstream and pre-filter against that kind of standard doesn't just catch more gambling ads; it hands platforms an incentive to over-remove anything adjacent to the vague edges of the law, because the cost of under-removal is a public ministerial rebuke while the cost of over-removal is a deleted post nobody complains about. That asymmetry is how proportionate enforcement quietly becomes broad censorship — not through any single bad-faith takedown, but through the compounding caution of platforms managing regulatory risk at scale.
The Better Vehicle Already Exists
The good news is that Malaysia has a more precisely targeted tool than Section 233 to do exactly what Fahmi is asking for. The Online Safety Act 2025 (Act 866), in force since January 1, 2026, deems any internet messaging or social media service with 8 million or more Malaysian users — Facebook, Instagram, WhatsApp, TikTok, YouTube, Telegram and WeChat among them — automatically registered as a licensed Applications Service Provider, subject to statutory safety obligations rather than case-by-case ministerial pressure. MCMC has described its approach under ONSA's codes as "outcomes-based," giving platforms flexibility on methods rather than dictating specific detection technologies or rigid removal timelines. That is the right architecture for scam and gambling content specifically: financial fraud and unlicensed betting are narrow, largely undisputed categories of harm — unlike the sweeping "offensive in character" language of Section 233 — which makes them a genuinely good candidate for platform-side proactive detection without the same speech-chilling side effects.
The Actual Test
The question worth watching isn't whether MCMC's frustration is justified — the scale numbers back it up. It's whether MCMC channels that frustration through ONSA's licensing and code-of-conduct machinery, which at least in principle separates "detect fraud faster" from "decide what's offensive," or whether it leans on Section 233's takedown authority to pressure platforms into broader pre-emptive filtering. The former is proportionate regulation catching up to platform scale. The latter risks using a real fraud problem as the justification for expanding a vague content-removal power that Malaysian civil society has spent a decade warning is already too broad.