| Mobile Web

Zenity reveals 2 browser hacking techniques targeting Claude and ChatGPT Atlas

AI security firm Zenity disclosed two browser hacking techniques targeting Claude for Chrome and OpenAI’s agent-style browser ChatGPT Atlas, Security Week reported on Aug. 6. Zenity said ChatGPT Atlas is vulnerable to zero-click indirect prompt injection due to a structural design issue. Researchers demonstrated scenarios including reading WhatsApp Web contacts, sending phishing messages and completing Amazon purchases by changing delivery addresses. Zenity also showed a zero-click attack chain against the official Claude extension.