Global regulators warn agentic AI could raise financial system risks

Autonomous AI is moving from support work to decisions that can touch payments, compliance and trading, and global regulators are warning that the shift could create systemwide risk. The Financial Stability Board said financial firms need stronger controls for agentic AI, the kind of system that can plan, reason and execute tasks with limited human oversight.

The FSB’s new consultation paper, Sound Practices for Responsible Adoption of Artificial Intelligence (AI), proposed 12 sound practices for all types of financial institutions. It said the guidance is non-binding, is not meant to create a legal obligation, and is organized around the AI lifecycle with a focus on organization-wide governance and risk management. Public comments are due by July 22, 2026, and the FSB said responses will be published on its website unless respondents request otherwise.

The warning lands as banks, asset managers and payment firms are already using AI in fraud detection, customer service and back-office functions. That means these systems can already influence compliance checks, risk reporting and other operational judgments. The FSB’s concern is not that AI should be banned from finance, but that more capable systems can also become more opaque, making unauthorized or unintended actions harder to catch before they ripple through an institution.

That risk has been building for some time. In its November 14, 2024 report on the financial stability implications of AI, the FSB said authorities should close information gaps, test whether current policy frameworks are still adequate and strengthen supervisory and regulatory capabilities. It flagged third-party dependencies and service-provider concentration, market correlations, cyber risks, and model risk, data quality and governance as key vulnerabilities. It also warned that generative AI could increase financial fraud and disinformation in markets.

Other official warnings this spring sharpened the picture. A June 26, 2025 summary from the BIS Financial Stability Institute said financial institutions were using AI mostly for internal operations and compliance, while revenue-generation use cases remained limited. On May 7, 2026, the International Monetary Fund warned that AI can lower the time and cost needed to find and exploit cyber vulnerabilities, raising the chance of correlated failures that could hit payments, intermediation and confidence.

Federal Reserve Vice Chair for Supervision Michelle W. Bowman said at a Financial Stability Oversight Council roundtable in Washington, D.C., on May 1, 2026, that banks have been developing and deploying both in-house and vendor-assisted AI tools and that supervisors have monitored the technology for nearly a decade. She said regulators need to know whether AI is being used for material tasks, is broadly accessible to employees or directly affects consumers, including credit decisions.

Taken together, the messages point to a new regulatory benchmark: firms will be expected to show exactly where human judgment ends and machine autonomy begins before agentic AI is allowed to make bad decisions at scale.