Deep Dive into Deepfakes

Anuj Gangahar, Editor-in-chief, Citi Institute talks to Rahul Sood, Chief Product officer from Pindrop at the Citi Future of Finance Forum 2025, focusing on the increasing threat of deepfakes. Sood outlines two main classes of deepfake attacks impacting financial services. 

Key topics covered:

• Types of Deepfake Attacks: Customer Impersonation, where fraudsters impersonate bank customers, often through call centers, leading to account takeovers and monetary fraud. The second type of deepfakes involve employee impersonation for fraudulent activities like money transfers and hiring scams. 
• Deepfakes Fuel Broader Societal Problems: Deepfakes spread misinformation, particularly on social media platforms with user-generated content, impacting elections and geopolitical events.
• The Scale of the Problem: The number of GenAI systems capable of voice and video cloning is rapidly increasing. The number of tracked deepfake systems has grown from 100-150 to 500 in the past year, largely due to the rise of open-source systems. These systems are becoming more accessible, cheaper, and require less data to create convincing fakes. Pindrop's analysis of 1.3 billion annual calls reveals a significant increase in machine-generated calls to call centers. By the end of 2024, over 2% of calls were machine-generated, a 6.5x increase from the previous year (less than 0.1%). This trend is expected to continue.
• Winning the Fight Against Deepfakes: While the rapid advancement of GenAI is concerning, Sood believes the fight against deepfakes is winnable. He argues that AI can be used effectively for deepfake detection.