The Financial Stability Board's FSB AI sound practices financial institutions 2025 consultation report, published 10 June 2026, proposes a menu of 12 sound practices designed to help boards and senior management govern AI adoption responsibly across their organisations. The report covers the full AI lifecycle, from development through deployment, and is structured around organisation-wide governance as well as stage-specific management. A public comment period runs through 22 July 2026, giving financial institutions, including credit unions, a narrow window to shape how these practices are ultimately finalized.
FSB AI Sound Practices: The 12-Practice Framework
The FSB AI Consultation Report situates its 12 sound practices within a pattern of escalating regulatory attention to artificial intelligence in finance. The Financial Stability Board has been building this body of work incrementally: a 2017 report addressed machine learning, a November 2024 report examined financial stability implications of AI broadly, and a 2025 monitoring report identified critical data gaps in how supervisors track AI adoption. The June 2026 consultation is the most operationally specific output yet. Rather than describing systemic risk in the abstract, the report proposes concrete practices that financial institutions could apply in their own governance structures. The framing is deliberately non-prescriptive: the FSB presents these as a menu, acknowledging that proportionate application will vary by institution size, complexity, and the nature of AI being deployed. Case studies drawn from real-world implementation are included to illustrate how proportionality can work in practice.
Generative and Agentic AI Get Specific Attention
One of the more forward-leaning elements of the consultation is its explicit treatment of generative AI and agentic AI alongside more established forms of machine learning. The FSB asks respondents directly whether the sound practices strike an appropriate balance between managing risks across all AI forms and addressing the specific risks that emerge from newer, more complex architectures. This matters because generative and agentic systems introduce risk profiles that differ meaningfully from classical supervised-learning models: outputs are less deterministic, human oversight is harder to embed, and vendor dependencies tend to be deeper. For credit unions that have begun piloting AI-powered tools, this distinction is not theoretical. Vendors are already shipping agentic features in workforce and member-service platforms. For a practical illustration of how that product layer is developing, see our coverage of Coconut Software's AI-powered branch workforce management solution, which captures how vendor-side AI is reaching frontline credit union operations. The FSB's consultation, by naming these AI categories explicitly, signals that regulators intend governance frameworks to keep pace with the product roadmaps of technology vendors.
What it means for credit unions operating in the $100M to $10B asset range
For credit unions in the $100M to $10B asset band, the FSB report lands at a useful moment. The National Credit Union Administration has not yet formalized an AI governance framework equivalent to what the Basel Committee on Banking Supervision is developing for larger banks, but the FSB's sound practices provide a credible reference point that NCUA examiners will almost certainly consult as they build out supervisory expectations. Credit union executives who wait for formal NCUA guidance before establishing AI lifecycle management protocols risk starting from a disadvantaged position. The FSB's 12 practices, applied proportionately, can serve as a working governance checklist right now. Key areas include board-level AI oversight, documented risk appetite for AI use, third-party and vendor risk controls, and ongoing monitoring of AI model performance after deployment. Institutions that have already invested in data infrastructure, as profiled in our look at AI governance and data platform capabilities at Skypoint, are better positioned to operationalize these practices without starting from scratch. The consultation's proportionality framing gives smaller institutions explicit permission to scope their compliance posture to their actual risk exposure.
What we're watching
- 22 July 2026: FSB public comment deadline on the AI sound practices consultation. Credit unions and their trade associations can submit responses via the FSB's secure online form; the FSB has specifically requested case studies from nonbank financial institutions.
- Q3 2026: Expected publication of an overview of public responses by the FSB, which will indicate whether the 12 sound practices will be narrowed, expanded, or restructured before a final report.
- NCUA supervisory guidance cycle: Watch for any AI-related examination guidance from the National Credit Union Administration in the second half of 2026, particularly any references to FSB or Basel Committee on Banking Supervision frameworks as supervisory benchmarks.
- No date for the final report is confirmed in the consultation document; compliance teams should monitor the FSB's website for updates on the final publication timeline.