One asset manager's AI tool turns risk reviews around in hours, but who signs off?
Boards at financial firms have a 12-point AI guide after the Financial Stability Board finalized its sound practices on October 15, 2026.
The guide is voluntary. The FSB calls the practices "non-prescriptive" and says they "do not constitute a binding international standard." There is no effective date. Firms get a toolkit they can tailor to their own context and risk profile, and boards and senior managers are strongly encouraged to reference it. The final version follows a draft released for comment on June 10, 2026.
From robo-advice to risk checks in hours
AI use in finance has picked up since the FSB's 2024 report on the topic, though the pace varies by firm, sector and jurisdiction. Surveys and industry engagement by authorities point to relatively advanced use at internationally active banks, large or sophisticated trading firms and asset managers, and payment providers.
Several financial institutions use AI to support robo-advisory services, from customer profiling and investment planning to portfolio management and rebalancing. Some large asset managers use AI tools to speed up compliance reviews of marketing materials.
One case study features a global asset manager whose AI tool reviews investment documents, checks proposals against approved mandate rules and produces an initial risk assessment within hours. A person still makes the call. The tool "does not approve or reject investments and final accountability remains with the risk manager." The FSB says its case studies are illustrations, not benchmarks.
The client risks worth a second look
The report also flags consumer protection and market conduct risks that may apply to investor protection as well. One item deals directly with advice: "Potential for mis-selling and unsuitable recommendations: The use of AI in customer-facing applications, including robo-advisory and product targeting, may lead to recommendations that are unsuitable or not aligned with customers’ risk profiles, particularly where models rely on incomplete, outdated or improperly calibrated data."
Leaning harder on AI in marketing, advisory and customer engagement may also blur who is responsible for what, the FSB says, which may make misconduct harder to catch quickly. And if firms don't clearly disclose AI's role, customers may struggle to understand or challenge decisions that affect them.
Autonomous AI agents get attention too. The report says they may take unauthorized or erroneous actions, and that sound practices specific to agentic AI and multi-agent systems "remain nascent."
Twelve practices, sized to fit
The first four practices cover governance across the whole organization. They call for boards and senior management to line up AI adoption with the firm's business model, risk appetite and strategy, along with clear roles, AI risks built into risk frameworks with proper documentation, and ongoing adaptation as AI evolves.
The other eight follow an AI tool from inception to retirement: assessing materiality and risk, selecting models, data governance, explainability and transparency, performance checks, human oversight, cyber and ICT risk, and third-party AI dependencies. The report says firms remain accountable for AI outcomes even when they rely on outside tools and services.
Size matters here. Large, complex or highly connected institutions may need more robust practices, while smaller firms and lower-risk uses may apply only the relevant ones or modified versions. The report's sources include publications from Canada's Office of the Superintendent of Financial Institutions, among them Guideline E-23 on model risk management. As the FSB puts it, "Applying the sound practices in this report also does not absolve a financial institution from its obligations to meet local legal and regulatory requirements related to AI adoption."
The full text of Sound Practices for Responsible Adoption of Artificial Intelligence (AI): Final report is available at https://www.fsb.org/2026/10/sound-practices-for-responsible-adoption-of-artificial-intelligence-ai-final-report/.
