On June 10, 2026, the Financial Stability Board published a consultation paper: "Sound Practices for Responsible Adoption of Artificial Intelligence." The comment period closes July 22. The final report publishes in October as a US G20 deliverable. It is not a rule. It is also not the kind of document community financial institution executives should file away and forget.
Here is what it says, why it matters, and what the timeline looks like.
What the FSB Is
The Financial Stability Board is an international body established after the 2008 financial crisis to coordinate financial regulation across G20 member nations. Its membership includes the Federal Reserve, the FDIC, the OCC, and the Treasury, alongside their counterparts from the UK, EU, Japan, Canada, and 20 other jurisdictions.
When the FSB publishes guidance, US regulators are in the room when it is written. That matters for how you should interpret the document's status: formally non-binding, practically influential.
What the Document Covers
The consultation paper proposes 12 sound practices organized around the full lifecycle of AI adoption at a financial institution, from how the board governs AI decisions down to how operational staff document what an AI system actually did.
The practices address six broad areas:
Governance and accountability. Who at the institution owns AI decisions? What does board-level oversight look like in practice? The document expects a named function, not just a policy that says AI is someone's responsibility.
Risk identification and management. How does an institution identify the risks introduced by an AI system before deployment? How does it track those risks over time as the model drifts or the regulatory environment changes?
Data quality and model integrity. AI outputs are only as reliable as the data they are trained on. The practices cover what institutions should be able to demonstrate about data sourcing, quality controls, and model validation.
Third-party and vendor risk. Most community financial institutions do not build AI, they buy it through their core vendor, their digital banking platform, or a point solution. The framework addresses what oversight looks like when the model sits inside a vendor's product, not your own infrastructure.
Explainability and transparency. Can the institution explain an AI-driven decision to a regulator, a customer, or a board member? The practices set expectations for what documentation of AI reasoning should look like. "The model said so" is not an acceptable answer in an examination context.
Human oversight and accountability. The document is consistent on this point: AI systems are tools, not decision-makers. Humans remain accountable for outcomes. The practices describe what that accountability chain should look like operationally.
Why Community Financial Institutions Should Pay Attention
The FSB does not regulate community financial institutions directly. So why does this matter?
Because the transmission mechanism from international standard to domestic examination expectation is well-established and moves in one direction.
US regulators who participated in writing this document will carry it into domestic rulemaking. The FFIEC, the interagency body that sets examination standards for community financial institutions, updates its handbooks when member agencies develop new frameworks. Its membership includes the NCUA, meaning credit unions fall within the same examination pathway. Guidance that feels abstract today tends to become a specific question during your next safety and soundness exam.
This is not speculation. It is already happening. The US Treasury published its own Financial Services AI Risk Management Framework in March 2026, explicitly aligned with NIST AI RMF components. The OCC has issued examination guidance referencing AI model risk, and the NCUA has addressed technology and AI-related risk management in guidance to federally insured credit unions. The FSB document represents the international coordination layer on top of a domestic stack that is already taking shape.
Community financial institutions that wait to engage until examiners are asking questions will be building governance frameworks under pressure. The ones that understand the direction now have a meaningful head start.
What to Do Right Now
The document suggests a useful self-assessment question: if an examiner walked in today and asked you to walk them through how your institution governs its AI tools, could you do it?
Specifically:
- Do you have a current inventory of AI tools in use, including those embedded in vendor products?
- Is there a named owner responsible for AI governance at the institution?
- Do you have a documented process for evaluating AI tools before deployment?
If the honest answer to any of those is "not really," that is worth knowing before the exam cycle, not during it.
The FSB document is publicly available at fsb.org. It is written for large international banks, so some sections will not translate directly to a community institution. However, the governance and accountability sections are the most portable.