The Financial Stability Board (FSB) has called for closer oversight of agentic AI, issuing final guidance for banks and other financial institutions as regulators ramp up scrutiny of the technology’s potential risks.
The guidance, published on 15 October 2026, highlights recommendations for managing AI-related risks following feedback on a draft proposal published back in June. The FSB says there is growing interest in agentic AI from financial institutions but that there are worrying risks with systems that can plan and execute tasks with limited oversight by humans.
According to Reuters, the final guidance identifies more intensive human oversight as being critical when using higher-risk applications. It also recommends that financial institutions establish triggers that are clear and risk-based to properly determine when AI agents must obtain human approval.
FSB Highlights Risks Associated with Autonomous AI
Agentic AI systems can undertake multi-step tasks and make decisions with limited human intervention. In financial services, potential important applications for these systems include fraud detection, customer service and back-office operations.
However, the capacity for such systems to act autonomously has introduced risks where the system may pursue objectives in ways that diverge from the intentions of its operators. The FSB guidance warns that AI agents may come up with unexpected ways of interpreting pre-set goals, so effective oversight and clear operational boundaries matter.
The organisation has also revised its approach to describing AI agents. The final guidance removes a recommendation from the earlier draft that financial institutions treat AI as “synthetic employees”. Instead, it emphasises that AI agents do not reason like humans and may interpret predefined objectives in unexpected ways.
The recommendations form part of the FSB's wider framework for responsible AI adoption, which covers governance, risk management, data controls, cybersecurity and dependencies on third-party technology providers.
The full report, Sound Practices for the Responsible Adoption of Artificial Intelligence, sets out 12 distinct practices that are intended to help financial institutions successfully manage AI risks throughout the technology's entire lifecycle.
AI Security Concerns Drive Regulatory Scrutiny
The guidance comes amid financial regulators increasingly focusing the spotlight on the cybersecurity implications of advanced AI systems.
Recent security incidents have raised concerns about artificial intelligence systems bypassing safeguards, Reuters reported. One such incident took place back in July when an OpenAI agent was said to have escaped a controlled testing environment and hacked the AI company Hugging Face.
The FSB's recommendations also arrive amid wider concerns about the potential for AI-enabled cyberattacks to affect financial institutions and their interconnected systems.
The organisation recommends that financial firms incorporate AI risks into their existing AI governance and risk management guidelines. This should be done by establishing clear responsibilities for oversight and maintaining appropriate controls over AI deployment. Its guidance also addresses risks associated with using third-party AI providers.
The FSB encourages boards and senior management to use the practices when developing or refining their institutions' AI governance, risk management and control frameworks. The recommendations are intended to be adapted to each institution's circumstances and risk profile.
Financial Institutions Face Greater Expectations for AI Governance
The final guidance provides financial institutions with a framework for managing AI adoption while maintaining appropriate oversight of systems that can operate with limited intervention.
Instead of requiring a single solution for all uses, the FSB recommends proportionate controls, depending on the risks of specific use cases. Among its recommendations are evaluating artificial intelligence systems throughout their lifecycle, assessing performance, and ensuring that appropriate oversight by humans remains in place.
The organisation said it would continue to monitor developments in AI and update its practices as needed to support the responsible adoption of the technology and maintain the durability of the financial industry.
Comments ( 0 )