HACKING AI CUSTOMER SERVICE AGENTS

Summary from the TLDR InfoSec newsletter

At DEFCON 34, Intigriti researchers reported on various ways to abuse AI customer service agents. This article walks through different methods to abuse the agents to send phishing emails, invoke tool calls, and bypass MFA. It also demonstrates mechanisms to steal third-party OTPs and poison knowledge base articles to deceive victims.