The Financial Stability Board published a consultation report in June 2026 distilling AI governance observations into 12 nonbinding "sound practices" for financial institutions. The recommendations consolidate governance expectations emerging across major financial regulators and are likely to influence supervisory practice for multinational institutions operating across jurisdictions.
Rather than creating new international standards, the FSB report distills governance practices already emerging across leading banks and insurers. It focuses on concrete governance actions covering the full AI life cycle: strategic direction, risk management, model selection, data governance, explainability, performance management, human oversight, cyber resilience, and third-party risk.
Use the 12 practices as a governance benchmark
Financial institutions should benchmark existing governance against the FSB framework, particularly for material, customer-facing, or third-party AI deployments. The recommendations align with existing regimes: the EU AI Act and DORA (board oversight, documentation, third-party risk) and U.S. financial regulators' focus on governance, explainability, data privacy, and consumer protection.
Firms with mature operational resilience and third-party risk programs can often extend those frameworks to AI rather than building entirely new governance structures. The FSB reinforces embedding AI governance into existing enterprise risk, compliance, operational resilience, cybersecurity, and third-party risk management.
What the case studies demonstrate
The FSB's case studies from banks and insurers show substantial productivity gains alongside governance maturity. Key themes: meaningful human oversight for material or customer-facing decisions; centralized AI inventories as adoption scales; continuous testing and monitoring; and enhanced vendor governance for third-party AI transparency and accountability gaps.
For example, a midsize bank applied generative AI to create credit narratives, reducing drafting time from over a day to 15 minutes, but with approximately 80% accuracy in instead of extraction - meaning about one in five extractions still required correction. A large internationally active bank built an in-house agentic fraud detection system that reduced fraud losses by more than 20% during the for the first half of the 2026 financial year versus the same period in 2025.
Practical steps for finance leaders
Finance leaders with board-level accountability should consider three concrete actions now:
- Benchmark existing AI governance against the FSB's 12 practices, particularly where AI supports material financial decisions - Maintain a centralized organizational AI inventory, supported by dedicated systems that track each use case throughout its life cycle - Strengthen third-party risk management to address the transparency gaps the FSB identifies - for example, a G-SIB piloted an integrated AI platform helping corporate relationship managers reduce proposal preparation time by roughly 50%, enabling three times more client dialogue by serving previously underserved clients
The AI Learning Path for CFOs offers practical governance guidance for finance leaders responsible for AI oversight. For broader coverage of AI in financial services, see AI for Finance.
The FSB's consultation period has closed. The final guidance may differ in material respects from the consultation version. Financial institutions should monitor final publication and continue assessing whether their existing governance aligns with the expectations reflected in the report.
Why this matters for finance professionals
Finance leaders are increasingly expected to demonstrate board-level understanding of AI risk appetite and oversight. The FSB framework provides a practical operating model that finance executives can use as a common global baseline while accommodating jurisdiction-specific requirements. For CFOs and financial controllers, the emphasis on centralized inventory governance, continuous monitoring, and documented accountability for third-party AI means that strong governance programs in these areas serve as a competitive advantage for adoption, not just compliance.
Your membership also unlocks:
