Regulators Push Banks to Tighten Controls on Autonomous AI Systems
The Financial Stability Board published recommendations Wednesday requiring banks to strengthen oversight of autonomous AI systems as their use spreads across financial operations. The guidance addresses what regulators see as a growing gap between the speed at which these technologies operate and the ability of humans to monitor them.
Autonomous AI-systems that make decisions and execute tasks with minimal human involvement-is already in use across banking. More than half of financial-sector organizations surveyed by the Cambridge Centre for Alternative Finance said they actively deploy agentic AI, with some moving beyond pilot programs to scale.
Banks use these systems for fraud detection, customer service, and back-office operations. The expansion has accelerated as more powerful AI models enter the market, creating what regulators describe as novel risks.
The oversight challenge
The FSB flagged a specific problem: AI agents can take actions that diverge from an organization's stated goals before employees notice. Unlike traditional software, these systems operate with enough autonomy that they can create problems across interconnected institutions before anyone intervenes.
Potential risks include unauthorized activities, security breaches involving sensitive data, and disruptions that cascade across multiple firms simultaneously. The FSB said in its report that "AI agents pose a distinct challenge for human oversight."
What regulators are asking for
The recommendations are voluntary and center on three areas: governance, human review, and operational safeguards.
- Banks should set clear limits on what autonomous AI systems can do
- High-risk decisions-particularly large financial transactions-should require employee approval before execution
- Safeguards should be built into processes from the start, not added later
The FSB also suggested treating AI systems as "synthetic employees" for governance purposes, adapting human-resources policies and internal controls to account for autonomous agents.
The proposals are open for public comment through July 22. The FSB has not indicated whether these recommendations will become mandatory requirements.
For finance professionals, the guidance signals that regulators expect institutions to move beyond treating AI as an experimental tool. The focus is shifting to how these systems integrate into existing risk and compliance frameworks.
Learn more about AI for Finance and AI Agents & Automation to understand how autonomous systems fit into financial operations and governance.
Your membership also unlocks:
