Japan's financial regulator tells firms not to hold back on AI for fear of risk

The Financial Services Agency (金融庁) published version 1.1 of its AI Discussion Paper (AIディスカッションペーパー) in March 2026. The first version came out in March 2025. The paper is in Japanese, and the translations here are ours.

The agency revised it after running the FSA AI Public-Private Forum (金融庁AI官民フォーラム) from June to December 2025. The forum found that AI use at financial institutions is advancing “at a very fast pace” (非常に速いペース). Customer-facing services, which “were hardly being offered” when the agency surveyed firms in November 2024, have reached the stage of limited launches or active planning.

Subject	The paper’s position
Its own status	It does not set supervisory expectations or required actions. It is a basis for dialogue
The risks it lists	It was not written to mean that a firm “must not introduce AI unless it has addressed every issue mentioned”
What it expects	Firms should “take on the challenge actively, without shrinking back excessively out of fear of risk” (リスクを恐れて過度に委縮することなく)
AI that recommends a product	Whether it is solicitation (勧誘) under the Financial Instruments and Exchange Act is decided on the facts of each case
Existing law	It applies whether or not a firm uses AI

The underlying survey ran from 3 October to 15 November 2024 and drew answers from 130 firms, about 40% of them deposit-taking institutions. The paper also cites a Bank of Japan survey from spring 2024 in which about 30% of institutions had started using generative AI, and about 80% were using, testing or considering it.

On customer service, most firms told the agency that hallucination makes direct use hard for now. Many already use generative AI to support call center staff, and the paper reports that new staff benefit most.

The paper compares AI with the internet, which was “hardly considered” for financial transactions thirty years ago and is now a core technology of finance.