GuardPhish

GuardPhish is a large-scale multi-vector phishing prompt dataset containing 70,015 samples from web, email, SMS, and voice attack scenarios, used to evaluate the susceptibility of open-source Large Language Models (LLMs) to adversarial phishing prompts.

Papers using GuardPhish (1)

GuardPhish: Securing Open-Source LLMs from Phishing Abuse2026