Malaysian financial AI adoption hits 70 percent amid governance push
BIS Speech

Malaysian financial AI adoption hits 70 percent amid governance push

Bank Negara Malaysia outlined governance priorities for artificial intelligence as domestic adoption surpassed 70 percent across financial providers. Speaking at the AICB Nexus Conference, the central bank mandated that boards retain ultimate accountability for automated systems.

Algorithms stop at the boardroom

More than 70 percent of Malaysian financial service providers have deployed at least one artificial intelligence application, driven by demand for efficiency and risk management.

Bank Negara Malaysia stated that rapid technological deployment requires institutions to move from internal operational gains toward collective defenses against fraud, scams, and cyber threats.

In addition, supervisory authorities affirmed that boardrooms and senior management cannot delegate legal and operational responsibility to automated models.

“Responsibility cannot be delegated to an algorithm,” the central bank emphasized, directing boards to place model explainability and risk appetite at the center of governance frameworks.

Phased data sharing by 2027

The central bank is preparing phased implementation of its Open Finance framework starting in 2027 alongside infrastructure partner PayNet.

Concurrently, asset tokenisation roadmaps are advancing to live pilots under the Digital Assets Innovation Hub.

These measures support the upcoming Financial Sector Blueprint 2027 to 2030, which establishes supervisory expectations around technology neutrality.

The Asian Institute of Chartered Bankers will support workforce transitions via its Future Skills Framework.

Governance cannot rely on good intentions

The central bank rightly places board accountability at the center of algorithmic risk governance.

Yet high-level appeals for ethical oversight remain ineffective without concrete audit standards and liability rules for opaque models.

Supervisors must enforce clear technical boundaries before operational dependencies become unmanageable.

Report an error