The Financial Stability Board (FSB), an international financial supervisor, maintains that artificial intelligence has the potential to profoundly transform financial services and entities' business models, but also to amplify or introduce new risks into the system that must be identified and managed appropriately.
In a report on best practices for the responsible adoption of AI, open for public consultation until July 22, the FSB emphasizes that this technology generates both opportunities and threats. Therefore, it urges financial institutions to understand and closely monitor the evolution of AI, as well as to respond with a clear strategy and robust protective measures to constantly changing risks.
Alongside the risks arising from inadequate or deficient use of AI, potential cyber threats, and high dependence on third parties due to the concentration of technology providers, the FSB's document focuses on risks linked to agentic AI, i.e., AI capable of operating autonomously, warning that this type of system can generate or intensify certain risks.
Among the threats associated with agentic AI, the report points out that AI agents can execute autonomous actions based on predefined objectives and also dynamically set or modify those objectives based on what they learn by interacting with their external environment, which could even generate the risk that they "undertake illegal, unethical, or unauthorized actions without human approval or oversight."
The FSB warns that these types of behaviors may only become visible when the agent is deployed in a real environment and that, once manifested, correcting or mitigating them can be very complex or even impossible for humans. In certain situations, effective supervision and detection might require the use of another AI agent.
"The responsible adoption of AI allows financial institutions to seize opportunities and benefits while minimizing associated risks," states the report, in which the FSB proposes a package of twelve best practice recommendations to guide the responsible implementation and development of this technology in the financial sector.
Among these guidelines, the body highlights the need for comprehensive AI governance throughout the organization, which will serve to decide whether or not to adopt a technology, how, and to what extent. In this context, it emphasizes the role of the board of directors and senior management in setting the overall approach and exercising oversight, so that the incorporation of AI is consistent with the entity's business model, risk appetite, and strategy.
In a second section, the supervisor insists on the importance of rigorously and systematically evaluating the risks of each AI use case from its conception, selecting AI models or systems appropriate to the objectives and operational and technical needs, and having robust data governance.
Finally, the FSB's recommendations highlight the need to specifically address the cybersecurity risks linked to AI, incorporating cybersecurity risk scenarios, promoting the exchange of relevant information with key stakeholders, and, where appropriate, using AI tools to manage these cyber threats. Likewise, the report calls for controlling the risks arising from the use of AI by third parties, with particular attention to performance, transparency, data quality, supply chain and concentration risks, and business continuity.