Cybersecurity · VIDEO
ChatGPhish: how Markdown manipulates AI assistants for advanced phishing
Permiso Security has discovered ChatGPhish, an attack that manipulates AI assistants through malicious Markdown. Attackers insert hidden code into web pages...
KEY TAKEAWAYS
- ChatGPhish uses Markdown for indirect prompt injections in AI assistants
- Malicious web pages generate deceptive responses with fraudulent links and fake warnings
- The technique collects user data and redirects to malicious sites via QR codes
- Security measures such as Markdown sanitization and link filtering are necessary
- User training is essential to recognize and counter these threats
