AI Regulation Tracker  /  Global standard-setters

FSB Opens Consultation on Sound Practices for Responsible AI Adoption in Finance

On June 10, 2026, the Financial Stability Board published a consultation proposing 12 sound practices for how financial institutions can adopt AI responsibly. Comments are due by July 22, 2026, and the FSB expects to publish a final report in October 2026. This is a draft out for comment, not a rule, and it binds no one.

By Anthony Guerriero  ·  The Leveraged Years Regulatory Research  ·  Published July 17, 2026  ·  Last updated July 17, 2026

On June 10, 2026, the Financial Stability Board opened a consultation on how financial institutions should approach artificial intelligence. The FSB is the international body, hosted at the Bank for International Settlements, that brings together central banks, treasuries, and market regulators from the major economies to coordinate on financial stability. It does not pass laws. It builds consensus that national regulators then choose whether and how to adopt. So when it publishes a draft like this one, the right way to read it is as a preview of the direction supervisors are leaning, put out for comment before anything is settled.

What the FSB actually proposed

The draft lays out 12 sound practices. The FSB groups them into three areas. The first four cover organisation-wide AI governance: the idea that a firm's board and senior management own AI risk the same way they own any other material risk, with clear lines of responsibility. Practices five through ten deal with managing and mitigating AI risk across the lifecycle, from how a model is developed and validated to how it is deployed and monitored once it is live. The last two, eleven and twelve, address AI-related cyber, information and communication technology, and third-party risk, which is the reality that most institutions are buying AI capability from outside vendors rather than building it in house.

The FSB frames the whole thing as optional. The report proposes what it calls a menu of practices that institutions could apply, not a checklist they must complete. It is written for boards and senior management to consider, not for a compliance team to file against.

On why the FSB is doing this at all, the report is direct. As it puts it, "The FSB's sound practices are designed to help financial institutions navigate their AI adoption responsibly in a rapidly changing technology landscape." That is the animating idea. Adoption is moving fast, the risks are real, and the FSB wants a common reference point that regulators and firms can talk about before national approaches drift apart.

What this is, and what it is not

I want to be careful here, because it is easy to see the FSB's name on a document about AI and assume a rule just landed. It did not.

This is a consultation. The document is a draft, out for public comment through July 22, 2026, with responses submitted through a secure form on the FSB site and most responses to be published afterward. After the comment window closes, the FSB expects to publish a final report in October 2026. Even that final report will be guidance from a coordinating body, not a binding standard.

The FSB is explicit about the limits. It says the sound practices are "not intended to establish an international standard, to impose a prescriptive approach for responsible AI adoption by financial institutions." Read that twice. The body itself is telling you this is not a standard and not a mandate. It imposes no disclosure obligation, no control requirement, and no filing duty on any institution in any country. Nothing in your legal responsibilities changed on June 10 because this draft came out.

What the draft does is set a reference point and shape expectations. When the FSB describes what responsible AI governance looks like, national supervisors read it, and over the following months and years pieces of it tend to show up in their own guidance and supervisory questions. That is the mechanism worth watching. It is a weathervane for where oversight is heading, not a wind you are required to turn into today.

Why this reaches US finance teams

The FSB is not a US regulator, and this consultation carries no direct weight over US institutions or filings. But the United States is an FSB member, and US agencies sit at the table. FSB sound practices have a habit of flowing into US supervisory expectations, showing up in the questions examiners ask and the frameworks that agencies point to when they describe good practice. That is the channel that makes this relevant well beyond the institutions the FSB formally coordinates.

For US CPAs, advisors, and finance leaders, the value is twofold. First, the substance travels. Board-level ownership of AI risk, validation and monitoring across the model lifecycle, and hard attention to vendor and cyber exposure are exactly the themes US regulators have been circling, so the FSB draft is a useful map of where the consensus is forming. Second, the timing gives you a heads-up. If a version of these practices becomes the common reference in October, the firms that already run their AI use against these three buckets will not be starting from zero. Treat it as a benchmark to pressure-test your own governance, not as authority for any change you make.

The practical thread is governance, and it is the same thread that turns into findings when it is ignored: who owns the AI decision, how the model was validated, who monitors it in production, and whether you can show your work on the data and the vendor behind it. The FSB draft does not order you to build those controls. It describes them as what responsible adoption looks like, and it tells you a room full of the world's financial regulators broadly agrees.

What to do now

Read the draft for what it is, a consultation with a comment deadline of July 22, 2026 and a final report expected in October 2026, not a rule that took effect. Map your AI use against the FSB's three buckets: governance and board ownership, lifecycle risk management, and cyber, technology, and third-party risk. Keep a named person accountable for every material AI use, especially anything touching customer decisions, risk models, or financial reporting. Write down your governance answers before an examiner asks: data provenance, validation, monitoring, and vendor oversight. If your institution is regulated abroad or has a view worth sharing, note that the comment window is open until July 22. And do not change a control or a disclosure on the strength of this draft alone. Base any change on your own applicable US standards and your regulator's and auditor's guidance.

Browse the full AI Regulation News tracker

Informational analysis for working professionals, not legal, accounting, or audit advice. Confirm how any standard or requirement applies to your situation with qualified professionals in the relevant jurisdiction.