Good Bot, Bad Bot: How Agentic AI Changes Fraud Detection

As agent-driven activity becomes more common, banks and merchants will need to distinguish legitimate AI agents from malicious bots. Traditional bot detection has focused on identifying automated activity, which fraud teams have typically treated as suspicious. That approach will no longer work when authorized agents begin acting on behalf of customers. 

This report examines how fraudsters can use AI and agents to commit identity fraud and scams, how fraud teams can use agents in detection and investigations, and what controls are needed to determine whether an agent should be trusted to act for a customer.

Key questions discussed in this Fraud Management report: 

• How will agentic AI change the way financial institutions detect and manage fraud?
• How can financial institutions distinguish legitimate AI agents from malicious bots?
• What controls are needed to determine whether an agent should be trusted to act for a customer? 

Companies Mentioned

Amazon, Arkose Labs, Bliink AI, DataVisor, FBI, Financial Stability Board, Fingerprint, Flagright, Hugging Face, Mastercard, NICE Actimize, OpenAI, Oscilar, Perplexity, Sardine, Target, Visa, Vouched