Financial institutions must remain accountable for use of third-party AI tools: MAS
New guidelines for AI use will be effective in phases starting from Oct 2027.
Financial institutions operating in Singapore must remain responsible for any unintended consequences arising from the use of third-party AI tools, according to new guidelines released by the Monetary Authority of Singapore (MAS) on October 7. These guidelines, set to take effect in phases starting from October 2027, aim to address the use of AI by financial institutions such as banks and insurers.
MAS emphasized that institutions are accountable for AI tools, regardless of whether they are developed, operated, or provided by third parties. To ensure safe usage, institutions are required to assess various aspects such as model transparency, explainability, and contingency plans before incorporating third-party AI tools into their services.
The guidelines also outline the need for human oversight, maintain inventories of AI use cases, and implement safeguards in a risk-proportionate manner. Basic AI governance policies, such as restricting the input of confidential or client information, are deemed sufficient when AI services perform poorly and do not significantly impact the company, its customers, or stakeholders.
The board and senior management are responsible for regularly reviewing AI-related policies and risks, establishing clear internal reporting processes for managing incidents, and ensuring an inventory of AI usage within the institution. This inventory should include details about the systems, models, approved modes of use, training data, and accountable personnel.
In addition, institutions must conduct risk assessments that consider potential consequences of poor AI performance, the extent of autonomy granted to AI systems, and the need for appropriate controls to mitigate harmful biases and discriminatory outcomes. These controls should be implemented according to the risk level, with stricter scrutiny and robust controls required for AI used in high-impact activities such as credit decisioning and insurance underwriting.
Financial institutions have a two-phase implementation schedule. By October 7, 2027, they must have foundational governance systems in place, including an inventory of AI tools and the ability to conduct risk assessments. By October 8, 2028, institutions must rigorously test, monitor, and maintain strict human control over their AI systems, while securing the necessary skilled staff and technology infrastructure to operate them safely.
MAS intends to further consult the financial sector in 2027 to determine additional guidance needed for agentic AI use.
Written by urgent.news from Straits Times's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
