The dual role of artificial intelligence (AI) in payments fraud highlights a profound dilemma. The same AI that can be exploited to perpetrate payments fraud can also be used to prevent and detect payments fraud.

Yet, in the paradox of AI, and the ongoing battle between the engineering and the prevention of payments fraud, lies the potential for AI to emerge as a sophisticated solution and dominant power to combat payments fraud, and ultimately tilt the scales in favour of AI as a stringent ally in fortifying payments security.

In recent times, generative AI and predictive AI used in fraud prevention systems have become important and frequently discussed topics among business leaders, finance, treasury, and information security professionals, given that fraud scams and bank fraud schemes caused US$485.6 billion in global losses last year, as outlined in Nasdaq and Verafin’s 2024 Global Financial Crime Report.

Generative AI is used to create or produce new content (text, image, code, audio, and video) from existing content or data, while predictive AI analyses current and historical data to make predictions about future events.

According to the  PYMNTS Intelligence Generative AI Tracker®: Can Generative Al Break the Payments Fraud Cycle?, “Generative AI could become a cornerstone of modern payments fraud prevention strategies, promising significant improvements in accuracy, efficiency and cost savings”, while predictive AI “Has achieved nearly prerequisite status in recent years for modern fraud detection.”

Generative AI and predictive AI serve distinct but complementary purposes in payments fraud prevention. Offering a comparative analysis, PYMNTS Intelligence examines generative AI versus predictive AI for various key applications in payments fraud detection and prevention, emphasising the notable strengths of generative AI. Here’s how they compare:

Detecting nuanced patterns

Generative AI is proficient at uncovering subtle, complex, and novel patterns that traditional rules-based systems fail to detect, like irregular or unusual transaction amounts and atypical spending activities, as per the PYMNTS Intelligence Generative AI Tracker®.

Conversely, PYMNTS Intelligence states that predictive AI leverages “A combination of historical data, statistical modelling, data mining techniques and machine learning (ML) to predict future outcomes” and flag suspicious transactions that deviate from the norm or established patterns. While it is adept at spotting known or familiar fraud scenarios, it might fail to catch novel or emerging threats.

Establishing baselines in payments activity

The Generative AI Tracker® by PYMNTS Intelligence reveals that while predictive AI utilises historical data to establish payments activity baselines, which necessitate periodic updates to ensure effectiveness, generative AI, in comparison, adapts continuously to changing payments behaviour, establishing dynamic payments activity baselines and “instantly identifying anomalies” such as “unusual login times and transaction patterns.”

Recognizing sophisticated fraud schemes

Generative AI demonstrates a distinct advantage over predictive AI in addressing sophisticated fraud activities. The PYMNTS Intelligence Gen AI tracker shows that, generative AI analyses large datasets encompassing both legitimate and fraudulent transactions to unearth complex fraud schemes spanning multiple accounts or channels.

On the contrary, predictive AI, processes large volumes of data efficiently to detect patterns suggestive of fraud, although it may require manual intervention to cope with new types of fraud.

Enhancing fraud prevention and detection by generating synthetic data and identities

Generative AI, as highlighted by PYMNTS Intelligence, produces synthetic payments data for model training, reducing dependence on real-world data, and mitigating privacy concerns while enhancing the precision and speed of fraud detection.

This is a significant area where generative AI holds a marked advantage over predictive AI, which because it depends on authentic historical data for training, may be constrained by privacy regulations and “Limited by diversity and variability of actual fraud data within these datasets”, as explained by the PYMNTS Intelligence Generative AI Tracker®.

Moreover, generative AI demonstrates a clear advantage in uncovering novel synthetic fraud. It recognizes patterns associated with fictitious identities and alerts on suspicious transactions by “Leveraging its ability to generate and compare synthetic identities”, the PYMNTS Intelligence Generative AI Tracker® additionally noted.

In contrast, predictive AI, efficiently detects fraud using established indicators but may encounter difficulties in spotting “novel synthetic fraud without retraining”, PYMNTS Intelligence further clarified.

Helping reduce false positives

When it comes to reducing false positives, generative AI has an edge over predictive AI. While generative AI, as per PYMNTS Intelligence, “Continuously learns and adapts, exhibiting a nuanced understanding of legitimate payments behavior” that lowers false positives and minimises purchase interruptions, predictive AI, on the other hand, can attain “low false positive rates with well-tuned models”, however it needs more regular updates to ensure ongoing “accuracy and efficacy”, the Generative AI Tracker® elaborated.

In conclusion, the variety and influence of generative AI use cases underscore the financial sector’s and its decision-makers’ enthusiasm for its potential to combat fraud, even as payments fraud evolves to become more intricate, sophisticated, widespread, and costly.

While predictive AI remains a valuable tool, current generative AI iterations are designed to complement it rather than supplant it. Nevertheless, as these technologies converge and their effectiveness improves, “The payments landscape will move closer to a watershed moment when the opportunity cost of even attempting payments fraud will likely far outweigh the potential gains”, the PYMNTS Intelligence Generative AI Tracker® reckons.

Given that highly skilled cyber criminals are actively targeting large companies worldwide and often manage to remain under the radar, CEOs, CFOs, corporate treasurers, and information security professionals must strengthen their defences.

Investing in AI-driven fraud prevention technology may be the game-changing solution that identifies and neutralizes payments fraud before it leads to further financial losses and reputational harm for organizations globally.