Skip to content

Sebenarnya.my (with AIFA chatbot)

FreeProfessional

Operator: sebenarnya.my | Type: Non-detector | Cost: Free

TL;DR

Government-run multilingual fact-check chatbot (MCMC); read with independence caveat. AIFA is Southeast Asia's only state-operated multilingual fact-check chatbot, covering Malay, English, Mandarin, and Tamil with documented reach of 70 million-plus views; the toolkit lists it as an alternative in 2B.1 because the Malaysian information environment cannot be described without it, but the toolkit's deployment recommendations sit with civil-society and newsroom partners, not with the operator.

What it does

Sebenarnya.my is a fact-check website operated by the Malaysian Communications and Multimedia Commission (MCMC) since 2017, with the AIFA chatbot launched under the AI untuk Rakyat initiative in April 2024. The chatbot accepts public queries about circulating claims, returns MCMC-validated fact-checks where they exist, and routes the lay public to the broader Sebenarnya.my archive. AIFA covers four languages (Malay, English, Mandarin, and Tamil), making it the only government-run multilingual fact-check chatbot in Southeast Asia and one of very few state-operated multilingual fact-check products globally that covers Tamil at all.

The platform's reach is the operational fact that puts it in the shortlist. The platform recorded 70 million-plus views since the 2024 AIFA launch, which is the largest documented public-facing fact-check audience in any of the six focus countries. For a fact-checker working on Malaysian content, AIFA outputs are part of the documentary record of what the Malaysian state has said about a given claim: useful evidence in either direction (matching corroboration or noted divergence) when stitching a verified position together. The platform does not itself produce synthetic-content forensics; image, audio, and video work routes to 1A / 1B tools.

When to use it

  • A fact-checker working on Malaysian content needs to check whether MCMC has already issued a state-validated debunk on a circulating claim, as documentary evidence about the Malaysian government's published position.
  • A regional newsroom partner wants to surface what Malaysian Tamil, Mandarin, and Malay-language audiences are receiving through the official channel, in order to compare against independent civil-society debunks.
  • An institutional research project on regional information environments needs the AIFA outputs as one of several data points on what state-operated fact-check infrastructure reaches the public in each country.
  • A fact-check coalition is producing a comparative piece on how state-operated and civil-society-operated fact-check channels handled the same claim cluster; the March 2026 Ramadan-aid King and Anwar deepfake cluster is the documented example.

Limitations

Limitations

  • Government-operated raises independence concerns. MCMC is the same body that holds enforcement powers under the CMA and the Online Safety Act 2025 ; AIFA outputs sit inside that institutional context.
  • The chatbot returns MCMC-validated fact-checks; the universe of claims AIFA addresses is shaped by MCMC's priorities, which may not align with what civil-society fact-checkers prioritise on the same content.
  • As a tipline-style intake, the response cycle is editorial and not real-time; users expecting a synthetic-content verdict in seconds will be disappointed.
  • Operational continuity is exposed to MCMC budget and policy decisions; the platform's architecture and reach can change with state direction in ways civil-society tiplines do not face.
  • Lao, Sinhala, Filipino, and Thai speakers are not served; AIFA's multilingual scope is Malaysia-internal (Malay, English, Mandarin, Tamil).

Privacy and threat model

The threat model that matters here is operator identity. AIFA is operated by MCMC, a Malaysian government body. A user forwarding or querying content through AIFA is sending content into a government channel, not a private channel and not a civil-society editorial workflow. That has real consequences in the Malaysian legal environment: Section 233 of the Communications and Multimedia Act remained a live enforcement tool through 2025–2026, the Online Safety Act 2025 came into force on 1 January 2026 with broad "harmful content" categories, and 3R (race, religion, royalty) enforcement continued to be applied to online speech. Content the user sends to AIFA is documentary evidence in that environment; the user should not assume it is operationally separate from the broader MCMC enforcement context.

For ordinary citizen queries on already-circulating public claims this is operationally adequate, and AIFA's 70 million- plus views suggest most use sits in that band. For source handling (a journalist's confidential informant, an activist sending content that could identify them, a politically sensitive whistleblower in the Malaysian legal environment), AIFA is not the right channel. Direct contact with an independent civil-society fact-checker, a non-Malaysian newsroom partner, or a regional fact-check coalition with explicit source-protection arrangements is the right path. The Malaysiakini January 2025 device-seizure and CMS-access incident documented in the Malaysia country page is the operational warning case for how MCMC and adjacent regulators can interact with journalistic systems in Malaysia.

The toolkit's editorial position, drawn from the broader scope of work, is that independent fact-checkers and civil society read AIFA outputs with the same scrutiny they would apply to any government source-of-record claim: useful documentary evidence, not authoritative verdict. Cross-verification through independent civil-society or newsroom channels precedes publication.

Source-protection override (S2 — state-linked or legally sensitive)

AIFA is operated by MCMC; any submission enters a Malaysian government channel that the CMA Section 233 and Online Safety Act 2025 enforcement regime can read. Mitigation steps:

  1. Consult Digital Safety, legal counsel, or the editor before submitting any politically sensitive content to AIFA, especially material concerning royalty, race, religion (3R), or named state actors.
  2. Do not route source-identifying material through AIFA at all; use the independent civil-society / regional coalition paths instead.
  3. If AIFA is needed as documentary evidence of a counter- claim or as a citable government readout, query a public claim only; never private or unpublished source material.

Source-protection override (S5 — private or encrypted group collection)

A user forwarding WhatsApp content into AIFA's intake exposes the forwarding chain to the operator. Mitigation steps:

  1. Use AIFA only for consented submissions or for claims already in public circulation; never for scraped or infiltrated private-group content.
  2. Where the case originates in a private WhatsApp or Telegram group, route through independent tipline partners (Indonesian Kalimasada, Thai Cofact, or a Malaysian civil-society contact) instead of AIFA.
  3. Treat any submission to AIFA as a disclosure event; record the chain of custody in the case file.

Country and platform applicability

  • Indonesia: not applicable as a primary intake channel. Indonesian content runs through Kalimasada and Meedan Check.
  • Laos: not applicable. AIFA does not cover Lao; the 2B.1 Lao tipline gap is structural across all entries in this cell.
  • Malaysia: primary deployment. Documented through the AI untuk Rakyat April 2024 launch; the March 2026 Ramadan-aid King and Anwar Ibrahim deepfake cluster was among the documented cases the chatbot processed alongside MCMC's broader debunk workflow.
  • Philippines: not applicable; Philippine content routes through #FactsFirstPH on Check (Meedan Check) and Viber/Messenger via SEEK / VERA Bot.
  • Sri Lanka: not applicable as an intake channel, but the Tamil-language scope of AIFA is the documented contrast point for the Sri Lanka 2B.2 Tamil tooling conversation. AIFA's Tamil UI is government-operated Malaysian Tamil; Sri Lankan Tamil fact-checking still leans on pan-Tamil tooling (X-CLAIM) with the Sri-Lanka- specific adaptation gap documented in LIRNEasia research.
  • Thailand: not applicable; Thai content routes through Cofact Thailand on LINE.

Platform applicability: AIFA is web-based and chatbot-style rather than tied to a specific messaging-app integration in the WhatsApp / LINE / Messenger sense; the user reaches it via the Sebenarnya.my web interface or the AIFA chatbot endpoint documented on MCMC channels.

A specific Decision 7 note: AIFA enters the shortlist via the B1 exception-pass framing recorded in the Selection Criteria. For most tools, English UI is operationally sufficient for the toolkit's audience of trained fact-checkers, and local UI is a bonus signal rather than a requirement. AIFA is the explicit exception: the audience is the Malaysian lay public, not trained fact-checkers, and Tamil, Malay, Mandarin, and English UI is operationally critical for that audience. The chatbot would not reach Malaysia's actual demographic mix without it. That makes AIFA the rare 2B.1 case where local UI is a selection criterion in its own right, not just a bonus.

How to access

Free public access. Citizens reach AIFA via the Sebenarnya.my website or the AIFA chatbot endpoint on MCMC's communications channels; the chatbot conversation is in any of Malay, English, Mandarin, or Tamil at the user's choice. There is no account creation, no install beyond the web interface, and no fee.

For a research partner or regional newsroom wanting to compare AIFA outputs against civil-society debunks, the Sebenarnya.my public archive is the documentary record. MCMC publishes methodology and operational details on its own channels; an independent assessment of those materials is part of the documentary research a comparison piece would draw on.

Cost (current as of 2026-05)

Free for the user. The platform is funded through MCMC operational budgets and the AI untuk Rakyat initiative. The hidden cost from the user's side is the operator-identity consideration recorded in the Privacy and threat model section, not a monetary cost.

Quickstart

(Quickstart for a citizen user; institutional partners considering a comparative-analysis workflow on AIFA outputs should adapt step 5 accordingly.)

  1. Open the Sebenarnya.my website or the AIFA chatbot endpoint published on MCMC channels.
  2. Choose the conversation language: Malay, English, Mandarin, or Tamil.
  3. Submit the suspect claim or query; text-based works most reliably; the platform is web/chatbot rather than messaging-app-native.
  4. Read AIFA's response as documentary evidence about MCMC's published position on the claim, not as a final independent verdict.
  5. For institutional or research use: cross-verify the AIFA output against an independent civil-society fact-check (Meedan Check coalitions; regional partners) before citing it in a publication; treat the AIFA reply as a documentary data point in a multi-source verification workflow rather than as a stand-alone authoritative source.
  6. For source handling, route through a civil-society fact-checker, a regional newsroom partner, or a coalition with explicit source-protection arrangements rather than through AIFA.

In the toolkit's workflow

Source-history pillar non-detector platform per Architectural Anchor 1. Sits in 2B.1 Multilingual tiplines as the Malaysia-anchored alternative alongside the global primary Meedan Check, the LINE-native Cofact Thailand, and the Indonesia-anchored Kalimasada. Of the four 2B.1 entries, AIFA is the only government-operated platform.

The toolkit documents AIFA because the Malaysian information environment cannot be described without it; 70 million-plus views and Tamil-language coverage are operational realities a Malaysia-relevant fact-check workflow has to account for. The toolkit's deployment recommendations, however, sit with civil-society and newsroom partners. What civil society and independent fact-checkers do with AIFA is observe its outputs as documentary evidence about the Malaysian information environment; what the toolkit recommends they deploy for their own intake is the civil-society alternatives elsewhere in 2B.1.

Standard combinations:

  • With Meedan Check at 2B.1 – the civil-society alternative; coalitions running Check-backed tiplines are what the toolkit recommends operators deploy.
  • With Kalimasada at 2B.1 – country-anchored civil-society contrast; same regional category, opposite operator type.
  • With Cofact Thailand at 2B.1 – civil-society contrast in a different focus country; volunteer-curated rather than state-run.
  • With MaLLaM (Mesolitica) at 2B.3 – Malay-specific open-source LLM; complementary for Malay-first verification phrasing on Malaysian content where civil-society Malay LLM output augments the cross-checking layer.
  • With Sinar Project iMAP at 2C.1 – independent network-monitoring layer; documented use includes detection of MCMC blocking events on competing outlets, and the iMAP card's framing belongs alongside AIFA as the independent-monitoring counterweight.
  • With Bernama MyCheck.My as a media-side Malaysian cross-check pointer (referenced in).

Decision-tree references: T7 tipline routing addresses Malaysian content with a primary node into AIFA at the documentary layer (the -ms row of the country and platform availability table) and a parallel routing into independent civil- society intake for source-handling and editorial work.

This card carries no detector signal class: AIFA produces a non-detector source-history signal (the MCMC-validated fact-check). Per Anchor 2 it combines with detector signals from Pillar 1 when both are present in the same case, but the operator-identity caveat above applies to how independent civil society reads the AIFA signal in any combined workflow.

Override notes

Override notes

  • Government-operated independence caveat (government-operated-independence-caveat): AIFA is operated by MCMC. The card carries this caveat in the TL;DR opening line, in the Limitations admonition, in the Privacy and threat model section, and in the In the toolkit's workflow section. Independent fact-checkers and civil society read AIFA outputs with the same scrutiny they would apply to any government source-of- record claim; the toolkit lists AIFA as documentary evidence about the Malaysian information environment rather than as a deployment recommendation for civil-society or newsroom partners.

  • B1 exception-pass for public-facing Tamil/Malay/ Mandarin/English UI (b1-exception-pass-Tamil-Malay- Mandarin-public-facing): AIFA enters the shortlist because its audience is the Malaysian lay public, where local UI is operationally critical rather than a bonus signal. The Tamil-language scope is the rare SEA government-run chatbot case to cover Tamil at all and is a documented contrast point for the Sri Lanka 2B.2 Tamil tooling conversation.

Sources

  • MCMC (Malaysian Communications and Multimedia Commission). Sebenarnya.my — fact-check portal and AIFA chatbot. MCMC, 2024. sebenarnya.my.