Financial Stability Board calls for tighter controls on agentic AI as adoption accelerates

Agentic AI Risk Catches Eye of Financial Stability Board | PYMNTS.com

In a report published Wednesday (June 10), the FSB said it was "strongly" encouraging these companies to consider establishing measures to offset risks from artificial intelligence (AI), including agentic AI systems. "The high levels of autonomy that AI agents may have can create or amplify certain risks, which can materialise at great speed," the report said. These risks include the threat that AI agents -- designed to act autonomously -- might take illegal, unethical or unauthorized actions without human oversight. "Overriding, redressing, or remediating these actions can be difficult or impossible for humans," the report said. In addition, AI agents could make incorrect decisions because of "goal misalignment, insufficient information and other reasons, such as reward hacking," the report added. Such actions may occur when the agent is deployed in a live environment, and could be difficult to monitor and detect in real time, the FSB said. To mitigate these risks, the FSB proposes a list of "sound practices" for financial companies to adopt, including monitoring AI adoption, and "adapting human resources controls and processes to AI agents in a way that treats them as synthetic employees." The FSB says it will take feedback on these non-binding guidelines until July 22. PYMNTS wrote last month that while agentic AI can lessen banks' burden when it comes to things like managing lending documentation or conducting compliance reviews, there are still challenges and liabilities related to delegation. "Automation allows a bank to accelerate work," that report said. "Delegation requires a bank to decide which responsibilities can be handed to software and under what conditions." PYMNTS cites the example of a fraud analyst using technology to spot suspicious activity, while an AI agent could put together account histories, cross-reference customer records, summarize findings and suggest next steps before human involvement begins. "The value proposition becomes apparent when viewed through the lens of productivity," the report added. "The governance challenge becomes apparent when viewed through the lens of accountability." The report added that financial institutions are dealing with this at a moment when risk management is becoming more demanding. Research by PYMNTS Intelligence and Block shows that 46% of financial institutions report increasing sophistication in fraud schemes, while nearly half of the executives surveyed point to regulatory pressures as a major challenge. "Those figures help explain why discussions surrounding agentic AI frequently return to governance rather than capability," PYMNTS added.

Global watchdog calls for tighter controls on agentic AI in finance

LONDON, June 10 (Reuters) - Global regulators said increasingly autonomous forms of AI could amplify risks for the financial system and called for new controls as adoption accelerates. The Financial Stability Board (FSB) in a report on Wednesday "strongly" encouraged boards to consider implementing safeguards to mitigate risks from AI, including from "agentic" AI -- or those systems capable of planning, reasoning and executing tasks with limited human oversight. o Agentic AI is already being used by financial firms for fraud detection, customer service and back-office functions. o 52% of financial sector respondents to the Cambridge Centre for Alternative Finance survey reported active agentic adoption, with 23% of them scaling or transforming and 29% piloting agentic functions. o Regulators and global standard-setting bodies have stepped up warnings about the risks posed by the rollout of AI across the financial sector since Anthropic released Mythos, viewed by experts as posing significant cybersecurity challenges to the banking industry. o The FSB, a global standard setter, said autonomous AI introduces risks that can "materialise at great speed", including the possibility of unauthorised or illegal actions, data breaches and disruption to connected systems. o "AI agents pose a distinct challenge for human oversight," the report said, warning they could pursue actions that stray from firms' intentions without staff being aware or able to intervene quickly. o To address those risks, the standard setter has outlined a series of proposed "sound practices", urging financial firms to define clear boundaries on AI use and embed safeguards. The non-binding guidelines are open for feedback till July 22. o They also include boundaries on what AI agents can do and require human approval for high-risk actions, such as financial transactions above certain thresholds. o Firms can also consider adapting HR controls and processes to AI agents in a way that treats them as "synthetic employees," the FSB said.