A global financial watchdog is urging banks to enhance their oversight of autonomous AI systems.
The Financial Stability Board (FSB) issued this recommendation in a report released on Thursday, October 15, which discusses appropriate practices for the adoption of artificial intelligence (AI).
According to the report titled “Sound Practices for Responsible Adoption of Artificial Intelligence,” “As adoption expands in scale and sophistication, AI may also amplify existing risks and introduce new ones, with implications for financial stability.”
It continues, explaining that less explainable models can erode transparency, while advanced autonomous AI systems might behave unpredictably or beyond intended limits. This could lead to unauthorized actions, data security issues, and disruptions to interconnected systems. Moreover, some sophisticated models may significantly increase the speed and magnitude of cyberattacks.
The report emphasizes that four of its recommended practices highlight the necessity for comprehensive AI governance within financial institutions leveraging this technology. It points out the vital role of the board and senior management in establishing an overarching approach and providing oversight to ensure that AI adoption aligns with the institution’s business model, risk appetite, and overall strategy.
The other eight practices provide a framework for managing AI-related risks from the beginning to the end of its lifecycle. Key areas of focus include systematically assessing the materiality and risks associated with AI use cases from their inception and continuously thereafter, as well as selecting suitable AI technologies and ensuring proper data governance.
The report suggests that responsible AI adoption can mitigate risks to financial stability, highlighting that financial institutions must also comprehend “the risks of not adopting AI, such as reduced competitiveness and a lesser ability to identify and handle cyber threats.”
Additionally, the FSB has retracted previous guidance issued in July, where it advised financial entities to treat AI as if they were “synthetic employees.” Now, the agency asserts that banks should recognize that “AI agents do not reason like humans” and might interpret set goals in unexpected ways.
It also notes, “The actions and capabilities of agents can change over time, especially if they gain access to new tools or receive new prompts.”
Recent research from PYMNTS Intelligence indicates that 95% of major financial services firms have broadly implemented or integrated newer AI tools within their data and technology processes.
However, there remains potential for greater integration. Another study from PYMNTS Intelligence shows that there’s a 25% adoption rate for AI in API orchestration and integration. Additionally, 30% of financial services firms surveyed cited data quality and fragmentation as their most significant barriers to further AI deployment.
