The Financial Services Agency of Japan finalised its Principles for Model Risk Management on 12 November 2021, setting eight principles for how Japan's systemically important banks govern every model they use, not only regulatory capital models. The FSA notes that models for AML, fraud detection and algorithmic trading increasingly draw on developments in 'machine learning and artificial intelligence', and its AI Discussion Paper of March 2025 points to the principles' risk-based approach and model inventory as the basis for managing AI model risk. The principles state expectations rather than prescriptive rules.
OFFICIAL TEXT: fsa.go.jp ↗ · IN FORCE · JAPAN FSA
| Document | Japan FSA Principles for Model Risk Management (2021) — Principles for Model Risk Management |
| Issued by | Financial Services Agency of Japan (with the Bank of Japan on frontier AI cyber measures) |
| Type | Guidance |
| Status | In force |
| Published | Nov 12, 2021 |
| Applies to | Japan's G-SIBs, D-SIBs designated by the FSA, and subsidiaries in Japan of foreign G-SIBs approved to use a model for regulatory purposes; the FSA may extend the scope |
| Official source | fsa.go.jp ↗ |
| Use cases | Model risk management · AI governance (general) · Third-party & vendor AI · AML / KYC |
What are the key points of Japan FSA Principles for Model Risk Management (2021)?
- Eight principles: governance; model identification, inventory and risk rating; model development; model approval; ongoing monitoring; model validation; vendor products and use of external resources; internal audit.
- A model is a quantitative process or system that turns data into outputs such as estimates, forecasts, scores or classification, including processes with qualitative or expert-judgement inputs.
- Scope covers all kinds of models, for example pricing, risk quantification, AML and market surveillance, whether or not they are used for regulatory purposes.
- Built on three lines of defence: model owners in the first line, an independent second line that validates models and can restrict or reject their use, and internal audit in the third.
- Vendor models: firms must manage risk despite limited access to proprietary methods and data, ask vendors for detailed information, validate on the best information available and keep contingency plans if a product becomes unavailable.
- Risk-based approach: the rigour and frequency of validation follow each model's risk rating, and firms should consider model risk in aggregate, including interdependence between models.
- The board and senior management own the framework, and overall model risk is reported to the board.
What did Japan FSA Principles for Model Risk Management (2021) change for banks?
The FSA had long required controls on internal models used for regulatory purposes; the principles extend model risk management to all models used by Japan's largest banks. The FSA's AI Discussion Paper (Version 1.0, March 2025) now points to them, especially the risk-based approach and the model inventory, as the starting framework for AI models.
Does Japan's FSA have model risk guidance that covers AI?
Yes. The FSA's Principles for Model Risk Management of 12 November 2021 cover all models used by Japan's systemically important banks, and the FSA's AI Discussion Paper (Version 1.0, March 2025) says it is appropriate to assess and manage the risks of AI models, including risks specific to AI, using the principles' risk-based approach, with the model inventory as one example.
Which banks must follow the FSA model risk principles?
Japan's G-SIBs, the D-SIBs designated by the FSA, and subsidiaries in Japan of foreign G-SIBs that have FSA approval to use a model for regulatory purposes. The FSA said it may widen the scope, and that its supervision looks at whether the framework works as a whole rather than checking each principle in isolation.
| Date | Document | Status |
|---|---|---|
| Oct 23, 2026 | BOJ Financial System Report (October 2026) — Financial System Report (October 2026) | Final |
| May 22, 2026 | Japan FSA/BOJ frontier AI request — Request regarding 'Short-term Responses by Financial Institutions in Light of Threat Changes from Frontier AI' (「フロンティアAIによる脅威変化を踏まえた金融機関等の短期的な対応」に係る要請について) | In force |
| Mar 3, 2026 | Japan FSA AI Discussion Paper v1.1 — AI Discussion Paper (Version 1.1): Preliminary Discussion Points for Promoting the Sound Utilization of AI in the Financial Sector | Final |
| Mar 4, 2025 | Japan FSA AI Discussion Paper v1.0 — AI Discussion Paper (Version 1.0): Preliminary Discussion Points for Promoting the Sound Utilization of AI in the Financial Sector | Superseded |
| Oct 4, 2024 | Japan FSA Guidelines on Cybersecurity for the Financial Sector (2024) — Guidelines on Cybersecurity for the Financial Sector (金融分野におけるサイバーセキュリティに関するガイドライン) | In force |
| Nov 19, 2026 | BNM AI Feedback Statement — Feedback Statement: Artificial Intelligence (AI) in the Malaysian Financial Sector | Final |
Follow every document these regulators publish
when one of these regulators moves, the next morning's brief says so · six sourced stories · 7 am PT · free
plus every regulation, bank and agent page update, the morning after · leave any morning