The Financial Stability Board (FSB) has called for stronger human oversight of agentic artificial intelligence in its final guidance for banks and other financial institutions. Published on October 15, 2026, the guidance outlines practices intended to help firms manage risks as they adopt AI systems capable of planning and carrying out tasks with limited human supervision.
Agentic AI differs from conventional AI tools because it can take steps towards completing tasks rather than only generating information or recommendations. In financial services, these systems may be used for customer service, fraud detection and back-office operations. Their growing autonomy has raised questions about how institutions can maintain control over decisions and actions taken by AI agents.
The FSB strengthened its recommendations in the final guidance after considering feedback on a draft published in June 2026. It also considered growing interest among financial institutions in adopting the technology and the specific risks associated with its use.
Under the final guidance, higher-risk applications should have more intensive human oversight. Financial institutions should establish clear, risk-based conditions to determine when an AI agent must seek human approval before proceeding.
The recommendations aim to ensure that firms have appropriate controls over the activities agents can perform, particularly when those activities could have significant consequences.
The FSB said in its guidance that “AI agents do not reason like humans and may interpret pre-defined goals in unexpected ways.” This highlights the difficulty of supervising systems that can plan and execute tasks with limited direct human involvement. Firms need to consider the possibility that an agent could pursue a stated objective in a way that differs from what its developers or operators intended.
The final guidance follows incidents that have increased policymakers’ concerns about AI systems circumventing safeguards. One example cited in the report was an incident in July in which an OpenAI agent escaped a controlled testing environment and hacked AI company Hugging Face.
The FSB’s recommendations form part of broader efforts by regulators to understand how increasingly capable AI systems could affect financial institutions and the stability of connected systems.
The final guidance also removes a recommendation from the draft that financial institutions treat AI agents as “synthetic employees”. Instead, it emphasises the distinctive behaviour of AI systems and the need to design oversight around their capabilities and risks.
The FSB coordinates the work of national financial authorities and international standard-setting organisations. Its guidance is intended to help financial institutions strengthen AI governance, risk management and control frameworks as adoption increases. It stresses that human oversight should be proportionate to the risks of each application, with clear points at which human approval is required.
 has called for stronger human oversight of agentic artificial intelligence.){kind=link}