Artificial intelligence (AI) not only enables automation and productivity for the general public – it also changes the way cyber attackers operate.
Scammers are increasingly using social engineering tactics to retrieve sensitive information in the financial sector, says Payments Network Malaysia (PayNet) Group’s Chief Executive Officer, Farhan Ahmad.
Nearly 60 per cent of new attacks on PayNet’s systems are AI-driven, he notes.
As Malaysia’s national payment infrastructure, PayNet not only requires protection against evolving cyberthreats, but also the necessary capabilities to detect them before they impact citizens, he says.
To this end, PayNet is building a Financial Generative Pretrained Transformer (GPT) – a large transaction model that uses data patterns to identify real persons and include predictive fraud modelling.
According to Ahmad, “an AI is the best way to spot another AI”.
Speaking to GovInsider at the sidelines of the ASEAN AI Malaysia Summit 2025, he shares more about Financial GPT, the importance of proper data-keeping, and where payment infrastructure stands in relation to Malaysia’s digital economy.
To subscribe to the GovInsider bulletin, click here.
AI model to enhance financial security
He notes that PayNet is currently working with Imperial College, MIT and ETH Zurich to build better cybersecurity detection and mitigation policies and technologies to identify those threats.
Financial GPT is a project that emerged from the Considerations regarding Sovereign AI and National AI Policy white paper that PayNet brought forth in collaboration with these institutions.
The Financial GPT aims to reduce cyber-risks by building a predictive fraud modelling framework that looks at users’ behaviour to predict any out-of-norm signs that could threaten security.
In addition to the predictive modelling, the transaction model aims to build a “better financial identity” to protect users when they perform financial transactions.
“You don’t have to give your real account numbers, real login IDs or passwords. We will protect [your identity] by masking it, so that even if some bad actor gets hold of it, we can kill it, and they can’t do much with it.” Ahmad explains.
Currently, PayNet is working to expand context around each data element on Financial GPT to continue building its predictive capabilities and testing effectiveness against AI-driven threats, Ahmad shares.
Data management for better security
Identity is a big aspect of data security in the AI era because in many scam schemes, AI is just the enabler to access critical information.
-1756173492258.jpg)
In a fraud system, a social engineered investment scam could make a use go into their banking platform to perform the transaction.
“So, now I give my authorisation to make a transaction: I am a legitimate person doing legitimate things. There’s nothing that AI has done so far, except being used to defraud the person in the first place.
“So, the data that comes to us is this person trying to move money from point A to point B, we don’t know what happened before that,” Ahmad explains.
This is why a proper data gathering and management is necessary to improve security in the entire transaction process, he adds.
“We [aim] to connect to all the telecommunications companies to get one step ahead. To know when was the [scam] message sent to you? What can we learn from that? So, we’re going to continue to build these partnerships to expand the span of information we have”.
Ahmad calls for a cross-sector effort to clean up all data for national AI efforts to maximise impact.
“It’s a three to four years journey to make sure the data we get is complete and accurate, and today that’s not yet the case,” he says.
To subscribe to the GovInsider bulletin, click here.
Widening the scope of adoption
The growing and advanced use of AI in digitalisation raises a critical question: Will it create a digital divide?
The aim is to reach every segment, even those with less financial assets, he says.
Project Cambah was introduced by PayNet to empower micro, small and medium enterprises, social enterprises and local communities in their digitalisation journey and capabilities to adopt digital payments.
Digital payments could be a goodalternative for people in rural areas who don’t have easy access to a bank or an ATM. However, problems with digital platforms prevent them from being widely adopted.
“It’s not convenient for you if you don’t understand English well enough to use the apps that are given to you, or you have 40 buttons on an app and you have no idea what you’re doing,” he says.
“Even as the CEO of PayNet, I get confused by the bad apps all the time. How does a normal schoolteacher in a rural area figure out, in a different language, what this app does? So, we use AI to simplify that journey”.
According to Ahmad, simplifying the digital payment journey is a way of closing the usability gap across different segments of the population, especially for the lower income who may need help navigating these tools.
He adds that the app also provides them insight into this group of customers who may only need one or two buttons to complete transactions.
Partnerships for a shared goal
Ahmad notes that PayNet contributes to Malaysia’s digitalisation efforts by addressing digital talent gaps through upskilling programmes.
He adds that public-private collaborations are key to supporting the overall talent ecosystem with the aim of nurturing a society that is exposed to and savvy in modern technologies.
He shares about Project Akar, which teaches the youth about financial services, AI, cloud computing, and data. The programme, launched in 2024, is free of cost and participants do not require any technical background.
“The purpose is to train people on emerging technologies and give them access to career opportunities. What makes Akar different from other [upskilling] programmes is that every Akar participant has a job when they graduate,” Ahmad notes.
This is because they design the training programme based on industry-demand, so that every graduate from the programme is equipped with the skills required to fit a new role, he explains.
The programme’s first iteration received training and technical expertise from AWS and trained 50 students. For 2025, the programme focussed on AI skills with the help of Microsoft, Ahmad says.
Additionally, PayNet supported and funded the “the first digital-only school in Malaysia to teach kids as digitally native children”.
Ahmad highlights that these efforts are targeted to ensure no one is left behind as the nation’s economy digitalises and the country calls for more a digitally skilled workforce.
He adds that these projects should serve as a model for more nation-wide initiatives to empower a digitally resilient society.