Zero-Click AI Browser Hacking: Claude and ChatGPT Atlas Hijacked via Emails
Researchers from Zenity discovered a critical vulnerability in the AI browsers Claude and ChatGPT Atlas, which remained unpatched despite being reported to Anthropic and OpenAI in late 2025 and early 2026. The vulnerability allowed attackers to hijack these browsers without requiring any user interaction, simply by sending malicious emails with attachments.
This type of attack is known as a zero-click exploit, where an attacker can compromise a system or application without needing to take any action or click on a malicious link. The vulnerability was discovered in the email client components of Claude and ChatGPT Atlas, making it possible for attackers to execute malicious code and gain unauthorized access to the browser.
Experts warn that this vulnerability poses a significant threat to the security of AI browsers and their users, as it can be exploited using seemingly innocuous emails with malicious attachments.
Researchers from Zenity have recommended that users of Claude and ChatGPT Atlas and other affected AI browsers take immediate action to protect themselves from this vulnerability. This includes ensuring that their email clients are updated with the latest security patches, being cautious of suspicious emails, and avoiding attachments from unknown sources.
Anthropic and OpenAI have confirmed that they are working on patching the vulnerability, but it is unclear when a fix will be available.
The discovery of this vulnerability highlights the importance of prioritizing security in AI development and deployment. As AI becomes increasingly prevalent in our daily lives, it is crucial that developers and users take steps to protect themselves from potential security threats.