Claude Extension Flaw Enabled Zero-Click XSS Prompt Injection via Any Website
Summary
A vulnerability in Anthropic's Claude Chrome Extension could allow any website to inject malicious prompts into the assistant without user interaction. This zero-click Cross-Site Scripting (XSS) flaw could enable attackers to trick users into executing unintended actions through Claude. The researchers highlighted that the vulnerability was present in the extension's handling of cross-origin requests.
IFF Assessment
The vulnerability allows for unauthorized prompt injection, which could lead to malicious actions being performed by the AI assistant on behalf of the user.
Severity
Defender Context
This highlights the ongoing risks of prompt injection attacks, especially when AI assistants are integrated with browser extensions. Defenders should monitor for vulnerabilities in AI-powered tools and educate users about the potential for such attacks. Prompt sanitization and robust input validation are crucial for AI models and their integrations.