What does Japan FSA AI Discussion Paper v1.0 say about AI in banking?

What are the key points of Japan FSA AI Discussion Paper v1.0?

• Chapter II: the paper does not express the FSA's monitoring viewpoints or the specific actions expected of institutions; it summarises use cases, challenges and governance examples from a survey of 130 firms (3 October to 15 November 2024) and interviews, and chapter IV 'should not be interpreted as requiring all mentioned challenges to be addressed before AI can be adopted'.
• Chapter I: the FSA's basic stance is technology-neutral and existing laws and regulations apply regardless of whether AI is used; it will make efforts to provide safe harbours by clarifying how regulation applies, and says it will review laws, regulations or guidelines if needed given AI's characteristics.
• Chapter III: more than 90% of surveyed institutions use conventional or generative AI in some way; use cases are grouped as internal use, indirect customer-service use and direct customer-service use.
• Chapter IV.1: challenges common to conventional and generative AI (data preparation, external vendors and risk management, return on investment); challenges made harder by generative AI (explainability, fairness and bias, development, operation and model risk management, personal information protection, information security and cybersecurity, specialist personnel); new challenges (hallucination, misuse for financial crime, financial system stability).
• Chapter IV.2: institutions' governance initiatives, including internal AI policies and rules.
• Chapter IV.3: demand for regulatory clarification ranked personal information protection, IT governance, model and risk management, and cybersecurity; the FSA will first consider revising principles and guidelines before legal responses, which it does not exclude where significant gaps exist.
• The FSA planned a Public-Private Stakeholder Study Group within the fiscal year (the FSA then held its AI Public-Private Forum from June to December 2025) and to update the paper as necessary; the paper also describes the Fintech Support Desk, which had handled 2,380 consultations from 2015 to December 2024.
• Chapter V describes the FSA's own use of AI in supervision.

What did Japan FSA AI Discussion Paper v1.0 change for banks?

It was the FSA's first consolidated statement on AI in finance and set a pro-adoption, dialogue-based tone: no new AI rulebook, existing laws apply, and the 'risk of inaction' is a stated concern. Banks use it to see how the FSA frames explainability, vendor risk, model risk and cyber risk for AI, even though it imposes no new obligations. Version 1.1 (3 March 2026) now supersedes it.

What did the Japan FSA's AI Discussion Paper Version 1.0 say financial institutions should do?

The FSA's AI Discussion Paper (Version 1.0), published 4 March 2025, imposes no new obligations: it says existing laws apply technology-neutrally, and that it does not state the FSA's monitoring viewpoints or specific expected actions. It expects institutions to identify and assess AI risks under a risk-based approach and with strong management involvement, to take on challenges proactively without being unduly deterred, to review business processes around AI, to develop use cases and to involve senior management. It organises the main challenges (data, vendors, explainability, bias, model risk, privacy, cybersecurity, hallucination, financial crime and stability) and records examples of governance, and the FSA says it will clarify regulatory application through dialogue and consider revising principles and guidelines first. Version 1.1 of 3 March 2026 supersedes it.

The paper is explicit that it is a foundation for dialogue. It says the FSA will explore the issues through ongoing dialogue and will take necessary actions such as updating the paper and clarifying the application of regulations. In the direction-of-response section, it says legal regulation should be limited to areas where self-regulation cannot be expected, so principles and guidelines would be revised first.

The survey base is voluntary and the paper declines sector-level analysis, so it should be read as a snapshot of practice at Japanese institutions in late 2024. The FSA's own later materials (Version 1.1) update the picture, noting rapid growth in customer-facing AI.

WHAT THIS MEANS IN PRACTICE

• Use the three-way classification of challenges as a framework for an AI risk assessment.
• Remember existing Japanese laws and supervisory guidelines apply to AI use; the paper adds no exemption or new rule.
• Document senior management involvement in AI strategy and risk decisions.
• Read Version 1.1, which supersedes this version, for the current FSA position.