Title: JPMorgan Chase Implements Advanced AI Models to Detect Fraud
In recent years, JPMorgan Chase has encountered an alarming surge in business-email compromise, which poses one of the most devastating types of attacks to both the bank and its clients. Ryan Schmiedl, the global head of payments, trust, and safety at JPMorgan, revealed that fraudsters target corporate entities due to their large number of employees who often have limited communication among themselves. As a result, the attackers find vulnerabilities within these organizations.
To combat this growing threat, JPMorgan has been leveraging large language models, similar to ChatGPT, to detect fraudulent activities and financial crimes. This adoption of advanced artificial intelligence (AI) technology reflects a broader trend embraced by various organizations, including banks, payments networks like Swift, and online gambling companies like Caesars Entertainment, as they seek to thwart malicious actors and suspicious transactions.
JPMorgan’s fraud detection technology has evolved from basic business rules to machine learning, and now, it relies on AI to extract entities from unstructured data and identify signs of fraud. For instance, large language models are employed to identify patterns and anomalies within emails. Schmiedl explained that there are unique patterns fraudsters utilize when crafting fraudulent emails, and through machine learning, these patterns can be learned and recognized.
The bank employs hundreds of AI-based models that evaluate various aspects such as behavior, payments, and new accounts, thereby effectively assessing risk. However, JPMorgan sources its training data solely from within its ecosystem to ensure that external data, which may not be audited and validated, does not compromise the accuracy and integrity of its results.
Besides JPMorgan, other industry leaders are also harnessing AI to enhance their fraud detection capabilities. Swift, for example, in collaboration with technology partners like Google and Microsoft, is integrating AI into its existing products and services, aiming to boost success rates in fraud prevention. By delving into their extensive historical data store of billions of transactions, Swift can identify anomalies and share these indicators with its bank members.
Similarly, Caesars Digital, responsible for the company’s online gambling platforms, is leveraging AI-powered fraud detection models to combat both first-party fraud (friendly fraud or account-owner fraud) and third-party fraud (hostile fraud). Through collaboration with external partners, Caesars Digital is developing and training AI models to effectively distinguish between legitimate and fraudulent activities.
While the adoption of AI technology holds immense promise in the fight against fraud, there is also concern among industry professionals. Schmiedl, Kelly (from Caesars Digital), and Bhatia (from Swift) acknowledge the potential for fraudsters to exploit AI technology for their own malicious purposes. This risk is driving JPMorgan to invest in innovative technologies and research to stay ahead of these adversarial attacks in the ever-evolving landscape of fraud prevention.
Addressing this challenge necessitates consistent investment, research, and effort. Although progress has been made in detecting deepfake voices and photos, there is an ongoing need for sustained advancements. As organizations continue to prioritize the development and implementation of advanced AI models, the battle against fraud will require ongoing vigilance and determination.
In conclusion, JPMorgan Chase’s utilization of advanced AI models, such as large language models, marks a significant step in the fight against fraud in the banking sector. With the integration of these models and ongoing investments in research, JPMorgan and other industry leaders are reinforcing their defense against fraudsters, safeguarding both their customers and their business operations.
