State Bank of Vietnam Seeks a Developer for an AI Credit Platform With a Vietnamese Finance LLM
The State Bank of Vietnam (SBV) is seeking an organisation to lead the research and development of an artificial intelligence platform that supports lending, with the aim of widening access to credit and improving risk management at credit institutions, Thoi bao Ngan hang (Banking Times) reported in an article republished by Vietnam.vn on October 22. The task was approved under Decision 2338/QĐ-NHNN as part of a special national science, technology and innovation programme for strategic technologies.
The order is detailed. The winner must build a credit knowledge data platform, a Vietnamese-language financial large language model, an OCR model to extract information from documents, a multi-agent orchestration platform, at least three AI agents for processing loan files and three models for credit prediction and scoring. The financial model must score at least 85 percent on the Vietnamese version of the MMLU benchmark, OCR accuracy must exceed 95 percent on financial statements and legal documents, and the scoring models need an AUC-ROC of at least 0.75 on independent test data, with false alarms capped at 20 percent.
There are fairness and outcome requirements too. The system must control bias across customer groups, explain the main factors behind a loan rejection or high-risk warning, and log its actions in real time. At participating credit institutions that use little or no AI today, electronically disbursed loans must rise by at least 50 percent after 24 months and processing time must fall by 40 percent, with results verified by an independent third party.
What it means in practice
The project would run for 48 months, from January 2027 to December 2030, with at least 40 percent localisation, at least one software copyright registration and one scientific paper on autonomous AI in financial risk management. Applications are due at the SBV's IT Department in Hanoi by 17:00 on November 10, 2026. For Vietnamese AI firms, the specification reads like a shopping list of products they have been pitching to banks, from Viettel's three AI platforms for banks to FPT's AI-first finance roadmap.
It also puts into practice the guardrails the central bank set earlier this month, when the State Bank of Vietnam set risk tiers and human review rules for banking AI. Explainable rejections and bias controls are exactly what those rules expect of high-risk credit models.
What to watch: which consortiums apply before the deadline, whether a technology group such as Viettel or FPT takes the lead, and which credit institutions volunteer for the trial.