PayPal has implemented a new AI-powered scam alert system to enhance fraud prevention for its global user base, including Venmo users in the United States, according to ZDNet. This initiative combats evolving scam tactics by leveraging artificial intelligence to identify and mitigate potential risks in transactions.
PayPal’s new system deploys AI-powered scam alerts specifically for Friends and Family transactions on the PayPal platform globally and for Venmo transactions within the U.S. These alerts notify users of a potential scam before money is sent. The system’s response varies based on the detected risk level, increasing friction as risk indicators rise.
Yigit Yildirim, SVP of global fraud prevention at PayPal, stated, “We need to be smarter and faster than the scammers, and we’re combating AI used by bad actors with more intelligent AI used for good.” He further elaborated on the system’s design, noting, “Consider this scenario we worked to avoid: If you send the same friend money for lunch every week but always receive the same generic warning to watch out for scams, it not only interferes with your payment experience, but it leads you to ignore the warnings altogether.”
Why Google Wallet is quietly cutting PayPal out
For instances indicating the highest confidence of a scam, the system will block the payment from proceeding. If fewer risky signals are detected, the system takes less drastic action, such as issuing alerts that encourage users to verify transaction safety or recommending specific actions. PayPal indicated that these differentiated alerts were intentionally selected to improve the effectiveness of fraud prevention, reasoning that consistent static warnings might lead to user disregard.
The system operates using continually learning AI models that adapt to new scam types. These models detect scams not explicitly trained on by analyzing billions of data points and updating themselves as patterns change. Yildirim explained, “With continually learning models, our AI can understand similarities between a known and a new scam and help us detect the new one. Backed by our risk controls and research, the AI is leveraged to alert customers to a potential scam, even if we have not seen that exact same scam before.” The model then stores this new information for future scam detection efforts.
