AI Regulation Tracker  /  Regulator guidance

Japan's FSA says AI advice keeps the suitability and explanation duties on the financial firm

The Financial Services Agency published Version 1.1 of its AI Discussion Paper on March 3, 2026, telling banks and wealthtech firms that existing conduct duties apply to autonomous AI advice. It is guidance, not a new binding rule.

By Anthony Guerriero  ·  Reviewed by The Leveraged Years Editorial Desk  ·  Published July 7, 2026  ·  Last updated July 7, 2026

Japan's Financial Services Agency has told the country's banks and wealthtech firms that handing investment advice to an AI system does not hand off the legal duty that comes with it. On March 3, 2026, the FSA published Version 1.1 of its AI Discussion Paper, a document it titles "AI Utilization in Financial Institutions and Initial Issue-Sorting toward Promoting Sound Utilization." The core message is direct: existing conduct duties apply regardless of whether a recommendation is produced by a person or by a machine.

What the FSA actually said

The paper is technology-neutral. Rather than write a separate AI rulebook, the FSA states that the obligations financial institutions already carry stay in force when AI enters the process. That includes the suitability principle (適合性原則), the duty of explanation accountability, and the customer-first conduct standard (顧客本位). A firm that lets an AI robo-adviser recommend a product to a customer remains answerable for whether that product was suitable and whether the customer received an adequate explanation.

Version 1.1 builds on Version 1.0, published in March 2025. The update is grounded in a survey of 130 companies and in the FSA AI Public-Private Forum, which ran from June to December 2025 and gathered financial institutions and other stakeholders to discuss AI use, risk management, and governance. Version 1.1 also adds a new section addressing AI agents, reflecting how quickly autonomous systems are moving from drafting text to taking actions.

The control expectations for autonomous advice

The FSA is specific about what it expects when AI gives autonomous advice. Firms are asked to control for the known weaknesses of these systems: uncertainty, hallucination, and opacity. The paper sets out practical measures. Firms should visualize the decision logic in plain language so a customer can understand why a recommendation was made. They should offer two-way chat rather than a one-way output, present alternatives instead of a single answer, and allow escalation to a human. On the governance side, firms should document the logic behind recommendations, use third-party review, and monitor with KPIs for skew in which products the system tends to recommend.

Read together, these are not abstract principles. They describe how a compliance team would evidence that an AI advice tool still meets the suitability and explanation duties a human adviser would have to meet.

What it does not do

This is where the paper needs to be read carefully. The FSA is explicit that the document is an initial issue-sorting exercise, not a binding regulatory-compliance demand. It does not create new penalties, does not set a compliance deadline, and does not reclassify AI advice as a separately regulated activity. The FSA frames it as a preliminary analysis and continues to invite stakeholder feedback as technology and business conditions change. The obligations it points to are the ones that already exist in Japanese financial conduct law. The paper clarifies how those obligations apply to AI; it does not add a new statutory layer on top of them.

That distinction matters for how a firm should respond. The right reading is not "wait for a binding rule," because the underlying suitability and explanation duties already bind. The paper is best treated as the regulator's current thinking on how it will apply existing conduct standards to AI-driven advice.

The cross-border angle

For a US reader, the FSA's approach is a deliberate contrast with two other models. The European Union's AI Act leans on classifying certain systems as high-risk, with obligations attached to that classification. The US SEC has proposed a predictive-analytics rule aimed at conflicts of interest in the technology firms use with investors. Japan is keeping AI advice inside its existing conduct-of-business framework instead of building a dedicated AI-advice regime. A wealthtech firm operating across these markets cannot assume one compliance posture satisfies all three. It needs to map how the same robo-adviser is treated as an existing-duty question in Japan, a risk-classification question in the EU, and a conflicts question under the US rule.

Browse the full AI Regulation News tracker

Informational analysis for working professionals, not legal advice. Confirm how any rule applies to your situation with qualified counsel.