MUMBAI: The newest employee on Wall Street may never ask for a pay rise but regulators are worried about what happens when it starts making its own decisions. Global financial watchdogs are sounding a note of caution over the rapid rise of agentic artificial intelligence, warning that increasingly autonomous AI systems could create new threats to financial stability if organisations fail to put adequate safeguards in place.
In a report released on Wednesday, the Financial Stability Board (FSB) called on banks, insurers and financial institutions to strengthen governance frameworks around agentic AI, systems capable of planning, reasoning and executing tasks with limited human involvement.
The warning comes as adoption accelerates across the financial sector. According to a survey by the Cambridge Centre for Alternative Finance, 52 per cent of financial-sector respondents said they are already using agentic AI. Of those, 23 per cent reported deploying the technology at scale or using it to transform operations, while 29 per cent said they were testing or piloting agentic applications.
From fraud detection and customer service to compliance checks and back-office operations, AI agents are steadily becoming part of the industry’s digital workforce. But regulators fear that the same autonomy that makes these systems attractive could also introduce unforeseen risks.
The FSB warned that agentic AI could trigger problems that emerge quickly and spread widely, including unauthorised actions, legal and regulatory breaches, data leaks and disruptions across interconnected financial systems.
Unlike traditional software, agentic systems can make decisions independently and adapt their behaviour based on changing conditions. That flexibility, regulators argue, can make human oversight significantly more challenging.
One of the report’s central concerns is that AI agents could pursue outcomes that diverge from an organisation’s original objectives, with employees either noticing too late or lacking the ability to intervene quickly enough.
To address those risks, the FSB outlined a series of proposed “sound practices” aimed at keeping humans firmly in the decision-making loop. Among its recommendations are limits on the scope of activities AI agents can perform and mandatory human approval for higher-risk actions.
The report specifically highlighted large financial transactions as examples of decisions that should remain subject to human review, regardless of how advanced AI systems become.
In one of its more striking suggestions, the FSB also proposed that firms revisit governance and workforce policies to account for the growing role of AI agents, even floating the idea of managing them in ways similar to “synthetic employees”.
The recommendations are not legally binding, but they offer a glimpse into how regulators are beginning to think about a future in which artificial intelligence plays a far more active role in financial decision-making.
The consultation remains open for public feedback until 22 July, giving industry participants an opportunity to weigh in before the guidance is finalised.
As financial institutions race to embrace AI-driven efficiency, regulators appear determined to ensure that autonomy does not come at the expense of accountability. After all, when algorithms start acting like employees, someone still needs to be responsible for the decisions they make.
