More than 70% of Malaysia's financial providers had adopted AI. What comes next?

Artificial intelligence had already moved beyond a small set of Malaysian banking experiments by the end of 2025. In its Annual Report 2025, Bank Negara Malaysia said more than 70% of financial service providers had implemented at least one AI application, with use growing in customer analytics, operational efficiency and risk management.[1][3]

That figure is an adoption signal, not a scorecard of customer benefit. It tells Malaysians that AI is increasingly present inside regulated finance, but it does not show how many applications reached production, how consequential their decisions were or whether every deployment produced a better outcome. BNM's next steps therefore matter as much as the percentage.[1][3]

The annual report's wording is precise: more than 70% of financial service providers had implemented at least one AI application. One provider running one limited internal tool and another operating many mature systems can both sit inside that headline. The number should not be read as 70% of decisions, customers or revenue being handled by AI.[1]

The underlying 2025 discussion paper gives useful context. Its survey covered 120 providers and reported adoption across banks, development financial institutions, insurers and takaful operators. It also found projects concentrated in customer analytics and marketing, internal operations, technology and cyber risk, and fraud or anti-money-laundering work.[3]

Many financial AI systems sit behind a service rather than on its front page. They may help flag unusual transactions, rank alerts for investigators, segment customers, support claims work, estimate risk or help staff find information. A customer can therefore be affected by AI even when no screen labels the interaction as artificial intelligence.[1][3]

The public-interest test is whether these systems reduce friction or losses without creating unfair treatment, opaque refusals or new security problems. BNM identified stronger fraud detection, better customer experience and improved risk management as intended value. Those outcomes still need measurement, monitoring and a route for people to correct errors.[1][3]

BNM issued its AI discussion paper in August 2025 and completed the public consultation in October. The annual report said a feedback statement would be released in 2026. It also pointed to an industry AI Governance Framework developed through the Chief Risk Officers Forum of the Asian Institute of Chartered Bankers with BNM's support.[1][3][4]

The framework's principles cover fairness, ethics, accountability, transparency, explainability, reliability and security. For institutions, these are not seven slogans to place beside a model. They translate into named owners, approval boundaries, documented data and tests, monitoring thresholds, incident handling and evidence that senior management understands the use case.[1][4]

A financial institution can start with a live inventory of AI-assisted processes and rank them by customer impact, autonomy, data sensitivity and reversibility. A fraud-alert prioritiser, a staff writing assistant and a system influencing a credit or claims outcome should not receive identical governance simply because all three use AI.[3][4]

BNM said it would support guidelines, best practices, knowledge-sharing and talent development. The useful evidence of progress will be whether providers can show reliable testing, human review where needed, meaningful explanations, supplier controls and measurable customer outcomes as systems move from pilots into larger operations.[1][4]