Unpacking the Promises and Pitfalls of AI in Cross-Border Payments

Phoram Mehta, Chief Information Security Officer and Senior Director, APAC, PayPal

The Asia-Pacific region is the world’s growth engine. By 2040, APAC will account for 42% of the world’s GDP[1]. Much of that growth will be fueled by innovation, and in particular new Artificial Intelligence (AI) technologies. In fact, businesses across the region will spend an incredible $78 billion on this between now and 2027[2]. We are at an inflection point; the AI breakthroughs of today are going to shape how we work and live for many years to come.

But the rapid rise of AI poses serious risks too. One of the most pressing challenges in e-commerce security today is the rise of cybercriminals leveraging AI to create sophisticated scams[3]. In a 2023 survey, 85% of cybersecurity experts polled directly attributed the rising number of cyberattacks to generative AI[4].

Unlocking the promise of AI in APAC requires businesses to address the region’s unique challenges. The digital economy in APAC is less mature than in markets like the US and EU, and trust is still a challenge for retailers in the region.

The region’s retailers must build that trust by balancing the incredible functionality that AI promises with safety and security for their customers.

Authenticating and authorizing payments with AI

Although AI poses some potential security risks for retailers, it will also play a central role in making payments safe. In fact, AI has already revolutionized the payment industry, where the growing volume of online transactions has long exceeded our ability to manually authenticate customers performing these transactions.

It can speed up KYC and AML checks, making efficient decisions about whether an extra layer of authentication is necessary[5]. This decision is an important one, as every extra verification measure, or payment failure is friction to the consumer.

Payment failures are a common occurrence and can stem from innocuous factors such as entering incorrect payment information. Companies like PayPal can combine machine learning with card, issuer, and merchant information to identify the best retry strategy and minimize issuer declines that interrupt user payment requests.

This improves merchants’ authorization rates, unlocking revenue for e-commerce businesses with increased successful transactions especially in a region where cart abandonment rates surpass the global average at 79%[6].

New innovations promise to bring even more convenience to consumers. Improved computing power and lower costs for data collection have empowered us to use AI to identify patterns within massive troves of transaction data to further improve security and customer experience[7].

Leveraging AI to combat fraud

Data is the best tool for balancing security and commercial needs, but only if businesses are willing to make the necessary investment. Getting the balance right is tricky, because by definition it means putting up barriers to lock out malicious actors, while ensuring those same measures don’t affect functionality for paying customers. 

But new tools are available to make the process simpler. For instance, PayPal’s security algorithms review purchasing histories against patterns of fraudulent behavior identified through analytics. AI can monitor potential threats round-the-clock and learn in real-time, even as cyberattacks grow more sophisticated.

In the past, training a deep learning model could take weeks or even months. However, advances in hardware and software have enabled us to retrain models quickly stay updated on changing patterns and fraud tactics.

Regionally, there is an extra complication, because Southeast Asia and India have high rates of unbanked or underbanked adults, who are unlikely to be using AI services, and who also represent a sizeable gap in data[8].

However, that appears to be changing. Research estimates that APAC’s digital payments growth rate will hit 19.8% by 2027, far outpacing Europe at 10.7% and North America at 6.5%[9].

As millions of people in APAC become active participants in the digital economy, using AI to ensure robust security measures built on quality data will be critical to combatting fraud and maintaining growth.

Caution must guide every decision

The growing excitement about AI is justified. But there is a difference between having a powerful tool and using it in a way that adds value both to your business and to customers. AI tools are typically quite complex and it is possible — even easy — to get an implementation wrong.

AI models should not be deployed prematurely, as poor-quality, underdeveloped data sets create service friction and are vulnerable to data poisoning. This can damage overall confidence in AI, which ultimately could mean fewer businesses and customers will be willing to use it.

Solutions for better implementation are available. In Singapore, PayPal’s Asia Pacific Cyber Defense Centre and Innovation Lab work tirelessly to provide security monitoring as well as test and evaluate the scalability of AI solutions for the wider region. Clear governance protects both businesses and users, and we prioritize responsible data science and work to ensure our innovations are deployed in compliance with each market’s data localization and security regulations.

AI is genuinely a transformative technology, and within a few years it is likely to be so ubiquitous that most vendors will be using it in some form to process payments.

Businesses will benefit most if they build security into their solutions, so that customers feel safe using it. That’s why it’s imperative that they understand both the benefits and the risks, and how to best minimize those risks.  



Contact Us - Filler feature box for Ad content
Media Resources - Filler feature box for Ad content