The Financial Stability Board on Wednesday published a consultation report setting out sound practices for the responsible adoption of artificial intelligence by financial institutions. The report proposes 12 sound practices spanning organisation-wide AI governance and lifecycle risk management.
In its release, the FSB said it strongly encourages the board and senior management of financial institutions to reference the sound practices as they consider business strategy, technology adoption and risk management in an increasingly AI-enabled environment. The FSB said the sound practices are not intended to establish an international standard or to impose a prescriptive approach on financial institutions.
TWELVE SOUND PRACTICES
The 12 practices set out expectations across three areas: organisation-wide AI governance (practices 1 to 4), the management of AI risks through the stages of AI development and deployment (practices 5 to 10), and AI-related cyber, ICT and third-party risks (practices 11 and 12). The first covers organisation-wide AI governance, including the roles and responsibilities of the board and senior management, alignment of AI activity with the institution's overall risk appetite, and integration into existing enterprise risk-management frameworks. The FSB says institutions should incorporate AI risk considerations into existing risk-management frameworks, such as model risk management, operational resilience and third-party risk, and enhance them or establish new ones where they do not sufficiently address AI risks.
The second area addresses the AI lifecycle, from inception, design and development through verification and validation, deployment, operation and monitoring, re-evaluation and retirement. The report says use cases should be managed at or throughout the different stages of the lifecycle so that deployments are supported by proportionate guardrails. The practices include evaluating the performance of AI use cases through performance assessment, testing and ongoing monitoring, including for data and model drift.
GENERATIVE AND AGENTIC AI IN SCOPE
The report explicitly covers generative AI and agentic AI, the latter referring to autonomous systems capable of planning, reasoning and executing complex high-level goals independently. The FSB said AI may amplify or introduce risks, citing among others model risk from the complexity and limited explainability of certain AI, cyber risks and third-party dependencies, and that some newer risks are specific to technologies such as agentic AI. The report also notes that providers of cloud infrastructure, hardware and AI foundation models are often highly concentrated and, in some cases, vertically integrated.
Some sound practices are complemented by case studies drawn from actual AI implementation practices by financial institutions, intended to illustrate how the sound practices may be applied in practice. Most describe banks and one describes an insurer; others cover public-private and supervisory initiatives such as the BIS Innovation Hub's Project Noor.
The FSB is inviting comments by 22 July 2026, and responses will be published on its website unless respondents request otherwise. The FSB plans to publish its final report in October 2026. The consultation document is available on the FSB website. The FSB said the sound practices build on and are broadly compatible with existing and ongoing work by the FSB and other standard-setting bodies, as well as by national and regional financial authorities. The FSB is chaired by Andrew Bailey, Governor of the Bank of England, and coordinates the work of national authorities and international standard-setters, bringing together national authorities responsible for financial stability in 24 countries and jurisdictions.