Global financial sector regulators have warned that increasingly autonomous artificial intelligence systems could heighten risks to the financial system, calling for stronger safeguards as adoption across the sector gathers pace.
In a report published on Wednesday, the Financial Stability Board (FSB), a global body mandated to work with local and international regulators to ensure financial stability, urged company boards to introduce measures to manage risks associated with AI, particularly “agentic” AI systems capable of planning, reasoning and carrying out tasks with limited human intervention.
Financial institutions are already deploying agentic AI for functions such as fraud detection, customer service and back-office operations. According to a survey by the Cambridge Centre for Alternative Finance, 52 per cent of financial sector respondents said they had adopted agentic AI, with 23 per cent using it at scale and 29 per cent running pilot programmes.
Regulators and international standard-setting bodies have intensified their scrutiny of AI in finance following Anthropic’s unveiling of its Mythos model, which experts believe could present significant cybersecurity challenges for the banking industry.
The FSB warned that autonomous AI systems could trigger risks that materialise rapidly, including unauthorised or unlawful actions, data breaches and disruptions to interconnected systems.
“AI agents pose a distinct challenge for human oversight,” the report noted, cautioning that such systems could take actions beyond a firm’s intended objectives before employees are able to detect or halt them.
To address these concerns, the FSB has proposed a series of non-binding “sound practices”, open for consultation until 2 July 2026. The recommendations call on financial firms to establish clear limits on AI usage, embed safeguards and ensure human approval is required for high-risk activities, including financial transactions above specified thresholds.
The watchdog also suggested that organisations could adapt human resources policies and governance frameworks to manage AI agents in a manner similar to “synthetic employees”.
