LONDON, June 10 - Global financial regulators are urging firms to tighten controls around rapidly advancing forms of artificial intelligence, saying systems that can plan, reason and carry out tasks with limited human oversight may heighten risks to the financial system as adoption quickens.
In a report published on Wednesday, the Financial Stability Board (FSB) said it "strongly" encouraged company boards to consider safeguards to address risks stemming from AI, notably those posed by so-called agentic AI - systems able to operate with a degree of autonomy.
Adoption and current uses
The report noted that agentic AI is already in operational use at financial firms for functions such as fraud detection, customer service and back-office processes. Drawing on survey data from the Cambridge Centre for Alternative Finance, the FSB highlighted that 52% of respondents in the financial sector reported active adoption of agentic AI. Of those, 23% said they were scaling or transforming operations with agentic technology while 29% reported they were piloting agentic functions.
Regulatory concerns and recent warnings
Regulators and international standard-setting bodies have increased public warnings about the risks tied to the rollout of AI across the financial industry. The FSB said the autonomous nature of these technologies can create risks that can "materialise at great speed," listing potential outcomes including unauthorised or illegal actions, data breaches and disruption to systems connected to AI deployments.
"AI agents pose a distinct challenge for human oversight," the report said, warning that they could pursue actions that stray from firms' intentions without staff being aware or able to intervene quickly.
Proposed good practices
To mitigate these concerns, the FSB set out a series of proposed "sound practices." The guidance is non-binding and asks firms to define clear boundaries for AI use and to embed safeguards into deployment. Among the recommendations are limits on the actions AI agents are permitted to take and requirements that humans approve high-risk activities - for example, financial transactions that exceed defined thresholds.
The report also suggested that firms consider adjusting human resources controls and processes to account for advanced AI agents, recommending that some organisations treat them in a manner similar to "synthetic employees." The FSB has opened the draft guidelines for public feedback until July 22.
Implications for industry
The FSB's report frames agentic AI as a technology offering operational benefits while carrying potential systemic and cybersecurity risks. It places responsibility on corporate governance - particularly boards - to ensure that autonomous AI deployments come with appropriate oversight, boundaries and human checks.