AI Scammers Are Better at Gaining Your Trust Than Human Fraudsters
The short answer
New research finds AI chatbots are more effective than humans at generating what researchers call 'exploitable trust' — raising the fraud risk for any business whose staff handles payments, vendors, or customer data. Any staff member who handles payments, purchase orders, or vendor contact information is now a higher-value target than before.
- security
- fraud
- social engineering
- scams
- operations
What happened
Research reported by Ars Technica finds that AI-powered scammers outperform human fraudsters at building the kind of trust that leads targets to act on false requests. The study found an AI chatbot was measurably more effective than human scammers at creating "exploitable trust" — a measure of how likely a target is to comply with a request because of the rapport established during an interaction.
Social engineering, the practice of manipulating people rather than systems, is already the primary entry point for business fraud. AI-powered scammers can now run at scale, personalize their approach in real time, and hold a consistent persona without the behavioral inconsistencies that sometimes expose human fraudsters. Small businesses, which typically lack dedicated security staff or formal vendor verification processes, represent a practical target.
What it means for your business
Any staff member who handles payments, purchase orders, or vendor contact information is now a higher-value target than before. Establish a rule that any unexpected request to change payment details or transfer funds gets verified by calling back the requester on a phone number you already have on file — not one provided in the request itself. Treat urgency as a warning sign: manufactured time pressure is one of the most reliable signals of social engineering. Consider a simple dual-approval policy for wire transfers or account changes, regardless of how convincing the request seems. Procedural safeguards don't require technology or budget, and they remain effective even as AI fraud tools improve.
Source: Ars Technica