As fraudsters get more sophisticated, credit bureaus are sharpening their tools to detect and curtail fraudulent transactions. It’s a concern that’s captured broad attention as popular peer-to-peer payments apps like Zelle and Venmo show security vulnerabilities. For many fraud victims, these transactions are irreversible, adding pressure on institutions to develop more robust solutions.

Credit bureaus, including FICO, are doubling down on machine learning and artificial intelligence to battle fraud. This week, FICO launched Falcon X,  technology that will enable data scientists at financial institutions to combine open-source machine learning capabilities with FICO’s machine-learning techniques to detect criminal activity.

Doug Clare, vice president of FICO, told Bank Innovation that the company will develop and install a fraud prevention AI toolkit for banks, but he emphasized that institutions also are deploying their own data scientists in fraud detection efforts. In addition, FICO launched a collaborative initiative among institutions called the Financial Crime Studio to support these endeavors.

Because real-time payment methods like Zelle, Venmo and others can’t be reversed, Clare emphasized that AI can play a useful role in detecting and addressing fraudulent transactions. The challenge is keeping the user experience simple with the added fraud detection capabilities, since customers tend to be less than enthusiastic about additional security questions or complicated authentication processes.

“It’s basically bringing the AI lab right into the deployment platform,” said Clare. “We’re giving bank the opportunity to monitor the performance of the AI that is deployed in the system, look at all the data that flows through the system and provide mechanisms for banks to incorporate their homespun AI and monitor the performance.” 

See also: Technical Shortfalls Leading to Higher Fraud in P2P Payments — at Venmo and Others

Despite the progress, finding a solution won’t be easy, since popular peer-to-peer services like Zelle and Venmo send payments via emails. Fraudsters can easily create new accounts with widely used email providers such as Gmail. 

“Criminals have exploited the rigid infrastructure that underpins our global financial system,” said Jason Keegan, who oversees the fraud line of business for FICO, in a statement. “This has allowed them to not only commit theft, but also finance drug trafficking, human smuggling and terrorist activity. With Falcon X, we set out to help institutions detect and prevent criminal activity before the real-time transfer occurs.”

AI can stop the fraud from occurring by setting up a challenge before the transaction is allowed to occur. This can be a notification, clarification request or an additional security question. If questions manifest during this process, the transaction is blocked.

AI, however, has not developed capabilities to retrieve stolen funds from accounts, Clare noted, therefore companies like FICO are developing tools to stop fraudulent transactions from occurring in the first place. As more traditional banks and payment platforms turn to same-day payment transactions, having flexible AI that can be applied to a variety of financial institutions is crucial to decreasing the rate of fraudulent transactions and stolen funds.

“We provide the tools, we provide the deployment framework, they bring their own smarts, they bring their own data science team in and it all it all kind of gels there,” Clare said.