Yudistira (MAFINDO)¶
Operator: mafindo.or.id | Type: Non-detector | Cost: Free
TL;DR
MAFINDO's centralised Bahasa Indonesia hoax-claim database and machine-to-machine API mapping circulating hoaxes against debunked claims. The country-anchored Bahasa Indonesia entry in 2B.2; the back-end claim-archive that Indonesian Check tiplines query when matching incoming forwarded content.
What it does¶
Yudistira is the structured hoax database MAFINDO maintains as a public reference and machine-readable API. A claim (typically the text or transcribed text of a forwarded WhatsApp message, TikTok caption, or Facebook post) queries the API and returns matches against MAFINDO's corpus of debunked claims. The output is "this claim has been debunked here" or "no match," not a de-novo fact-check. Yudistira sits in the same MAFINDO product family as ASE (Anti-Hoax Search Engine, the public-facing search interface) and Kalimasada (the WhatsApp Hoax Buster bot, the citizen-facing intake on Meedan Check); Yudistira is the database layer all three products draw on.
The tool's value to the Indonesian fact-check ecosystem is specificity: MAFINDO's debunked-claim corpus is curated in Bahasa Indonesia by Indonesian fact-checkers across the CekFakta consortium, which means the matching is on the language and the claim universe Indonesian misinformation actually occupies, not a multilingual fallback retrofitted to the country. For a cross-coalition deployment that needs authoritative Indonesian debunk lookup, Yudistira is the documented data source.
When to use it¶
- An Indonesian fact-checker or newsroom needs to check whether a forwarded WhatsApp claim has already been debunked by the CekFakta consortium before opening a fresh verification cycle.
- A Check-backed tipline operator (Kalimasada, a CekFakta partner outlet, a regional Indonesian newsroom) needs the MAFINDO database integrated into the matching layer for duplicate detection.
- A regional research project on Indonesian information flows needs programmatic access to the curated debunk corpus for pattern analysis or cross-country comparison.
- A regional partner outside Indonesia is processing Bahasa-language content from Indonesian diaspora or cross-border discussions and needs the authoritative Indonesian debunk reference.
Limitations¶
Limitations
- Highly localised; requires continuous manual updating
. The corpus is editorially curated by MAFINDO and CekFakta partners; new claims must be added manually as they are debunked. - Bahasa Indonesia only. Filipino / Tagalog, Thai, Malay, Sinhala, Tamil, and Lao speakers are not served; for multilingual content the matching path is Meedan Alegre (XLM-R fallback) or the appropriate country-specific entry. - As a database, Yudistira returns matches against previously debunked claims; novel claims that have not yet been processed through MAFINDO's editorial workflow return "no match" and require fresh verification rather than being flagged as suspicious. - The corpus reflects MAFINDO's editorial priorities and coverage; claims that fall outside CekFakta's operational focus (very local rumours, fast-moving content not yet escalated to MAFINDO) may not be represented.
Privacy and threat model¶
The privacy posture is database-query rather than source-handling. A user or system queries the API with a claim text; the API returns matches. The query content travels to MAFINDO's infrastructure and is processed there. For ordinary duplicate-detection use this is operationally adequate; the query is typically the text of an already-circulating public claim, not source-identifying material.
For source handling (sensitive content from a specific informant, a clip that could identify someone, a private document), Yudistira is the wrong path. Direct contact with a MAFINDO journalist or a CekFakta partner newsroom with explicit source-protection arrangements is the right path; the database query layer is for circulating public claims, not for private source material.
The Indonesia legal environment context applies broadly: ITE risk is reduced after the April 2025 Constitutional Court rulings but not removed, and MAFINDO operates inside that environment. Yudistira itself does not change that framing; operational source-protection discipline at the editorial layer is the toolkit's recommendation.
Country and platform applicability¶
Decision 7 framing applies: Yudistira is an NLP-class tool, and content-language coverage is the operational selection criterion. Bahasa Indonesia is the documented operational language; the tool does not cover other SEA languages.
- Indonesia: primary deployment. Documented use through MAFINDO's Global Hoax Database and the ASE Anti-Hoax Search Engine; integrated with the CekFakta consortium's editorial workflow and queryable from Kalimasada-routed Check tipline content. The 2024 election cycle and the January-February 2025 Prabowo / Sri Mulyani deepfake cluster are the documented case anchors .
- Laos: not applicable. No Lao coverage; the 2B.2 Lao gap is structural and the supporting Lao path remains Google Cloud Translation routing (2A.1 cross-link) plus partner-mediated review.
- Malaysia: not applicable as a primary tool. Malay content is linguistically adjacent to Bahasa Indonesia but the corpus is Indonesian rather than Malaysian; Malaysian content routes through Alegre and the 2B.3 LLM layer (MaLLaM).
- Philippines: not applicable. Filipino / Tagalog content routes through Alegre.
- Sri Lanka: not applicable; Sinhala content routes through Dissect and Tamil content through X-CLAIM.
- Thailand: not applicable; Thai content routes through Alegre.
Platform applicability: not platform-specific; Yudistira is a back-end database queried via API. Its outputs feed whichever front-end intake (WhatsApp, Facebook, TikTok, LINE) the operating organisation runs.
How to access¶
Free public access via API and web. The MAFINDO Global Hoax Database is published on MAFINDO's channels alongside the TurnBackHoax.id public archive and the ASE Anti-Hoax Search Engine. Institutional partners considering programmatic integration should contact MAFINDO directly to confirm current API access conditions and any operational constraints on high-volume querying.
Cost (current as of 2026-05)¶
Free for the public user. Underlying operational costs sit with MAFINDO and are funded through MAFINDO's grant stack and CekFakta consortium contributions; the hidden cost is the continuous manual editorial work that keeps the corpus current.
Quickstart¶
(Quickstart for an institutional partner integrating Yudistira into a fact-check workflow; lay users typically reach the corpus via ASE or Kalimasada front-ends.)
- Confirm Bahasa Indonesia is the relevant content language for the case; if not, route to the appropriate 2B.2 entry.
- Submit the claim text to the Yudistira API endpoint published on MAFINDO's developer channels.
- Review the API response: matched debunks (with links to the verified fact-checks) or no-match.
- For matched debunks, route the link to the editorial layer so the operator's response references the existing MAFINDO verification rather than initiating a duplicate cycle.
- For no-match results, route into the operator's standard fresh-verification workflow: Kalimasada queue if the intake is WhatsApp; CekFakta partner outlet desk if the case warrants newsroom investigation.
- For high-volume programmatic use, coordinate with MAFINDO on rate limits and any usage-attribution requirements before sustained integration.
In the toolkit's workflow¶
Source-history pillar non-detector tool per Architectural Anchor 1. Sits in 2B.2 AI claim extraction as the country-anchored Bahasa Indonesia entry alongside the Sri-Lanka-specific Dissect, the multilingual Meedan Alegre, the English baseline ClaimBuster, and the Tamil pan-Indian X-CLAIM. Cross-cell into 2B.1 is operational: Yudistira is the MAFINDO debunked-claim corpus that Indonesian Check tiplines (Kalimasada at Kalimasada (MAFINDO)) query underneath the Alegre matching engine.
Standard combinations:
- With Kalimasada (MAFINDO) at 2B.1 – same maintainer, native integration; Kalimasada-routed forwarded WhatsApp content matches against the Yudistira corpus on the way through the Check editorial workflow.
- With Meedan Check at 2B.1 – Check is the workspace; Yudistira is the MAFINDO data source the Indonesian Check deployment queries.
- With Meedan Alegre at 2B.2 – Alegre is the multilingual matching engine; Yudistira is the curated Indonesian corpus Alegre matches against in the Indonesian deployment. The two cards describe the engine layer and the data layer of the same Indonesian back-end.
- With MAFINDO Satgas Pemilu at 2A.4 – Satgas Pemilu is the workflow design pattern that orchestrates Yudistira lookups during election cycles; the MAFINDO 2024 Election Task Force ran on this combination.
- With ASE Anti-Hoax Search Engine as the public-facing search interface MAFINDO operates against the Yudistira corpus (referenced for completeness; ASE is not in the 2B.2 shortlist as a primary tool).
- With InVID-WeVerify at 1B.1 – when the matched claim involves image or video content needing forensic verification before the editorial reply.
Decision-tree references: T7 tipline routing carries Yudistira as the Indonesian debunked-claim corpus underneath the Kalimasada / Check tipline node; cross-cell routing for direct institutional research access is via the documented MAFINDO API channels.
This card carries no detector signal class: Yudistira produces a non-detector source-history signal (database match against the curated debunk corpus). Per Anchor 2 a database match is one signal class supporting the editorial decision; the matched fact-check itself is the next signal class.
Override notes¶
Override notes
- Non-detector declaration (g2-non-detector): Yudistira returns database matches, not detector signals. Render in workflow as the Indonesian Bahasa-specific corpus layer underneath the country's tipline and matching infrastructure.
Sources¶
- MAFINDO. Yudistira — Global Hoax Database and Anti-Hoax Search Engine. MAFINDO, 2024. mafindo.or.id/yudistira.
- MAFINDO. TurnBackHoax.id — Anti-Hoax Search Engine. MAFINDO, 2024. turnbackhoax.id.