T&S Asset Management
Listen日本語

Rules on AI in lending, market by market

What regulators require when a lender uses AI to approve, price or manage loans: when each rule applies and who it covers. Southeast Asia, with Hong Kong, the EU and the United States for reference.

Last checked 22 Oct 2026

Vietnam

Regulation (binding)

State Bank of Vietnam, Circular 53/2026/TT-NHNN on AI systems in banking (issued 30 Sep 2026)

Applies from
15 Nov 2026; AI already in use must comply by 1 Mar 2027
Who is covered
Credit institutions, foreign bank branches, payment intermediaries, e-money providers, credit information companies and others

What is required when AI is used in lending

  • Class every AI system as high, medium or low risk before use; AI that proposes credit decisions is high-risk from VND 50 million
  • High-risk AI: impact assessment covering fairness, explainability and discrimination, and board approval before launch
  • Staff can override, pause or stop the AI; a contested automated decision is reviewed by a person
  • Customers are told when they are dealing with AI
Source

Singapore

Guidelines (supervisory expectations)

Monetary Authority of Singapore, Guidelines on AI Risk Management (issued 7 Oct 2026)

Applies from
Oversight and risk management from 7 Oct 2027; life-cycle controls from 7 Oct 2028
Who is covered
All financial institutions regulated by MAS, for all AI

What is required when AI is used in lending

  • Board and senior management accountable for AI risk
  • An inventory of AI systems and an assessment of how material each use is
  • Data governance, testing, human oversight, monitoring and change management across the life cycle
  • The institution stays accountable for AI supplied by third parties
Source

Thailand

Guidelines (voluntary)

Bank of Thailand, AI risk management guidelines for financial service providers (issued 12 Sep 2025)

Applies from
No date stated in the published summary
Who is covered
Financial institutions, specialised financial institutions and payment service providers, for in-house and third-party AI

What is required when AI is used in lending

  • Board and senior management accountable; an AI policy and named owners
  • Human oversight scaled to risk, stronger for customer decisions such as loan approval
  • Explainable outcomes, with inputs, outputs and parameters documented
  • Customers are told when they interact with AI and can reach a person
Source

Indonesia

Guides (minimum reference for banks; reference for fintech associations' codes)

OJK, Indonesian Banking AI Governance guide (Apr 2025); OJK and fintech associations, AI code of ethics guide for fintech (24 Nov 2023)

Applies from
No date set
Who is covered
Commercial banks; fintech including online lending (pindar)

What is required when AI is used in lending

  • Human oversight treated as an absolute requirement
  • Board oversight and an AI committee (legal, compliance, risk, data, security, customer service)
  • Controls from design and testing to periodic evaluation and audit
  • An OJK official said on 17 Sep 2026 that AI credit scores should not be the sole basis for lending
Source

Philippines

Principles (voluntary, not binding)

Bangko Sentral ng Pilipinas, Memorandum M-2026-031, Governance Principles for Artificial Intelligence in Financial Services (24 Jun 2026)

Applies from
No date set
Who is covered
Banks and non-banks supervised by BSP, and their AI vendors

What is required when AI is used in lending

  • Governance in proportion to the risk of each AI system
  • Guard against bias and discrimination; monitor after deployment
  • Documentation so that outputs can be understood and audited; people stay responsible
  • Tell customers, in product terms or other documents, when AI is part of a product or process; offer opt-in or opt-out where applicable
Source

Malaysia

Discussion paper (consultation closed 17 Oct 2025)

Bank Negara Malaysia, discussion paper on AI in the Malaysian financial sector (5 Aug 2025)

Applies from
Not a rule
Who is covered
Banks, insurers and e-money issuers

What is required when AI is used in lending

  • Sets out BNM's proposed approach to responsible AI, technology-neutral and proportionate
Source

For reference

Hong Kongreference

Supervisory guidance

HKMA, Supervisory Guidance for Responsible Adoption of AI, under the 'ART' framework (15 Oct 2026)

Applies from
No date stated in the announcement
Who is covered
Banks

What is required when AI is used in lending

  • Built on accountability, resilience and trust
  • A discussion paper on supervising AI agents (registry of material agents, intervention) is to follow
Source

European Unionreference

Law (binding)

AI Act, Regulation (EU) 2024/1689, Annex III point 5(b); dates amended by Regulation (EU) 2026/1744

Applies from
2 Dec 2027
Who is covered
AI that assesses creditworthiness or sets credit scores of individuals, placed on the EU market or used in the EU (fraud detection excluded)

What is required when AI is used in lending

  • Treated as high-risk: risk management, data governance, technical documentation, record-keeping
  • Human oversight, accuracy, robustness and cybersecurity
  • Lenders using it must assess the impact on fundamental rights
Source

United Statesreference

Law (binding)

Equal Credit Opportunity Act, Regulation B (12 CFR 1002.9)

Applies from
In force
Who is covered
Creditors in the United States, whatever model they use

What is required when AI is used in lending

  • An applicant who is declined must be given the specific principal reasons
  • Saying the applicant failed to reach the score, or did not meet internal standards, is not enough
Source

General information only, not legal advice. Rules change: check the current official text and take professional advice before acting.