Phishing Scams Can Deceive Large Language Models | Security Magazine

Netcraft researchers found that large language models (LLMs) provide incorrect login URLs one-third of the time, with 30% leading to unregistered domains and 5% to unrelated organizations. This poses a significant risk as threat actors can register these domains and set up phishing sites. Security leaders emphasize the need for guardrails to validate domain ownership and real-time URL validation to mitigate this risk.