eKYC Document Extraction

Vision LLMs for OCR and key information extraction from ID documents.

1in Indonesia

Named Models

Grab custom vision LLM 1BGrabReleasednot in the notesRuns onnot in the notesClassOCR and documentGrab's eKYC document extraction vision LLM serving live production traffic (post 4 Nov 2025); within about 3 pts of the 2B model.Commercial use: not stated; Jakarta hosting: unknown; Indonesian support: listed by developer

Grab custom vision LLM 2BGrabReleasednot in the notesRuns onnot in the notesClassOCR and document

Earlier 2B Grab eKYC vision LLM (full fine-tune): Thai +70 pts, Vietnamese +40 pts vs baseline.Commercial use: not stated; Jakarta hosting: unknown; Indonesian support: listed by developer

LayoutLMMicrosoft AIReleasedDec 2019Runs onnot in the notesClassOCR and documentMicrosoft document AI models pretrained jointly on text, layout and image for forms, receipts and invoice understanding.Commercial use: not stated; Jakarta hosting: unknown; Indonesian support: not stated

Where this comes from

Models the research names for this use case. Research, with sources on each model record.

Who Runs It

production, 2

PT Bank Tabungan Negara (BTN)Country: IndonesiaBale by BTN e-KYC with OCR, face recognition and liveness; also staff knowledge management, transaction anomaly detection and AI document processing for credit (in development)Vendor: not statedproductionDate1 Jul 2026None statedEvidence: company statementConfidence: lowSource: keuangan.kontan.co.id

GrabCountry: multi-countryGrab custom vision LLM about 1BVendor: GrabproductionDate4 Nov 2025Serving live production traffic; covers Bahasa Indonesia, Thai, Vietnamese, EnglishEvidence: company statementConfidence: lowSource: engineering.grab.com

Where this comes from

Documented uses by named organisations. The evidence column says who made the claim.

Candidate Models

Real5-OmniDocBench: PaddleOCR-VL-1.6 paper table

A general benchmark for this model class, not a test on this use case. Document parsing on real-world captures. score. Higher is better.

PaddleOCR-VL-1.6Baidu vendor-reported93.19

PaddleOCR-VL-1.5Baidu vendor-reported92.05

GLM-OCRZ.ai (Zhipu) vendor-reported90.32

MinerU2.5-ProOpenDataLab vendor-reported88.96

dots.ocrrednote (Xiaohongshu) vendor-reported86.38

MinerU2.5 (MinerU-2.5)OpenDataLab vendor-reported85.61

DeepSeek-OCRDeepSeek vendor-reported73.99

DeepSeek-OCR 2DeepSeek vendor-reported73.01

050100

PaddleOCR-VL-1.6Baidu, OCR and documentReleasedJun 2026runs on edgehas scoreslast 18 months

GLM-OCRZ.ai (Zhipu), OCR and documentReleasedJan 2026open weightshas scoreslast 18 months

DeepSeek-OCR 2DeepSeek, OCR and documentReleasedJan 2026open weightshas scoreslast 18 months

DeepSeek-OCRDeepSeek, OCR and documentReleasedOct 2025open weightshas scoreslast 18 months

MinerU2.5 (MinerU-2.5)OpenDataLab, OCR and documentReleasedSep 2025open weightshas scoreslast 18 months

dots.ocrrednote (Xiaohongshu), OCR and documentReleasedJul 2025open weightshas scoreslast 18 months

Unlimited OCRBaidu, OCR and documentReleasedJun 2026open weightslast 18 months

PP-OCRv6Baidu, OCR and documentReleasedJun 2026open weightslast 18 months

Nemotron OCR v2NVIDIA, OCR and documentReleasedApr 2026open weightslast 18 months

Falcon-OCRTechnology Innovation Institute (TII), OCR and documentReleasedFeb 2026open weightslast 18 months

Showing 10 of 73. All models of this class

Where this comes from

Derived, not researched: models in the directory whose class matches a class this use case uses, excluding family records and models shut down or deprecated. Ranked by how many fit facts they meet, then by release date. Each fit fact is read from the model record.

Use Case

Industry: finance and insurance
Model classes: OCR and document
Benchmarks or datasets: not in the notes
Regulatory notes: Surat Edaran Menkominfo 9/2023 Etika Kecerdasan Artifisial, Kementerian Komunikasi dan Informatika (now Kementerian Komunikasi dan Digital)eKYC document extraction is an AI-based service run by an electronic system operator, and the circular asks such operators to build and use AI on the values of inclusivity, transparency, humanity and security.
UU 27/2022 Pelindungan Data Pribadi, DPR and President of the Republic of Indonesia (administered by Kementerian Komunikasi dan Digital)eKYC document extraction processes biometric data such as face images or voice recordings of the people being identified or served, so the controller needs a lawful basis, must protect the data and must honour data subject rights.
Tata Kelola Kecerdasan Artifisial Perbankan Indonesia (2025 guideline), Otoritas Jasa Keuangan (OJK)eKYC document extraction is an AI use in banking, which OJK's AI governance guideline covers across the AI life cycle, including bias, explainability and cyber risk.
Panduan Kode Etik Kecerdasan Buatan yang Bertanggung Jawab dan Terpercaya di Industri Teknologi Finansial (2023), Otoritas Jasa Keuangan (OJK) with the fintech associationseKYC document extraction is also offered by fintech providers, for whom OJK's AI code of ethics sets principles of being beneficial, fair and accountable, transparent and explainable, and robust and secure.
POJK 22/2023 Pelindungan Konsumen dan Masyarakat di Sektor Jasa Keuangan, Otoritas Jasa Keuangan (OJK)eKYC document extraction deals directly with financial consumers, so the provider's duties on fair treatment, clear information, data protection and complaint handling apply.
POJK 8/2023 Penerapan Program APU, PPT dan PPPSPM di Sektor Jasa Keuangan, Otoritas Jasa Keuangan (OJK)eKYC document extraction supports customer due diligence or transaction monitoring, which financial service providers must run under their anti-money laundering and counter-terrorist financing programme.
UU 8/2010 Pencegahan dan Pemberantasan Tindak Pidana Pencucian Uang, DPR and President of the Republic of Indonesia (PPATK as financial intelligence unit)eKYC document extraction helps identify and report suspicious transactions, a duty this law places on reporting parties.
EU AI Act, Regulation (EU) 2024/1689 (international), European UnioneKYC document extraction uses biometric identification, which the EU AI Act lists among high-risk uses; it applies when the system is offered in the EU market.
UU 24/2013 Perubahan UU Administrasi Kependudukan, DPR and President of the Republic of Indonesia (Ditjen Dukcapil, Kementerian Dalam Negeri)eKYC document extraction matches customers against population identity data, which this law governs as population data managed by Dukcapil.

Where this comes from

From the industry tracks T08 to T10 of the research notes.

R&D Topics

Links to R&D topics are awaiting review.

Where this comes from

Which R&D topics would most change this use case. Drafted by Claude from the topic definitions and the use case; proposed until approved.

Incidents

Deepfake photos used to bypass digital KYC in Indonesian loan applicationsIndonesia, 2024-12-04, financialGroup-IB reported that one Indonesian financial institution client saw over 1,100 attempts to pass its digital KYC for loan applications with AI-generated deepfake photos.
Hong Kong Syndicate Allegedly Used AI-Generated Facial Composites to Open Bank AccountsHong Kong, 2025-04-07, securityHong Kong police reportedly arrested eight individuals accused of using purportedly AI-generated facial composites to open bank accounts with altered ID photos.
Deepfake ID Sales Allegedly Persist as OnlyFake Reportedly Relaunches with New Fraud Tools2024-03-01, securityResearchers from Au10tix discovered the relaunch of OnlyFake, a site offering AI-generated fake IDs.

Where this comes from

Documented failures, harms or misuse in this kind of use, Indonesia first, then the region, then well-known cases elsewhere (AI Incident Database, OECD AI incidents monitor, reputable news). Not a claim about any deployment listed here.

In Neighbouring Countries

Bank of Ayudhya (Krungsri)Thailand, production, 2019-03-06Facial recognition identity verification against the ID card photo for account opening

Where this comes from

Documented deployments of this use case in Malaysia, Singapore, Thailand, Vietnam and the Philippines, each with its source.

Opportunities

Opportunities are awaiting review.

Where this comes from

Gaps found by rule in what the directory holds for this use case. Each is a proposal for review.

Links

IndustryFinance29 use cases, 46 deployments

Use CaseeKYC Document ExtractionFinance

Models

Grab custom vision LLM 1BGrab, OCR and document

Grab custom vision LLM 2BGrab, OCR and document

LayoutLMMicrosoft AI, OCR and document

Who Runs It

PT Bank Tabungan Negara (BTN)In production, 1 Jul 2026

GrabIn production, 4 Nov 2025

Graph

Models 3

Grab custom vision LLM 1BGrab, OCR and document

LayoutLMMicrosoft AI, OCR and document

Grab custom vision LLM 2BGrab, OCR and document

Deployments 2

GrabGrab custom vision LLM about 1B, multi-country PT Bank Tabungan Negara (BTN)Bale by BTN e-KYC with OCR, face recognition and liveness; also staff knowledge management, transaction anomaly detection and AI document processing for credit (in development), Indonesia

Sources

How we built a custom vision LLM to improve document processing at Grab engineering.grab.com https://engineering.grab.com/custom-vision-llm-at-grab

To cite this page: Project AURION, AI Use-Cases Research & Innovation, "eKYC Document Extraction", https://aurion.semai.id/usecases/fin-ekyc-document-extraction/, records of October 2026.