Technology newsroom
🚨 Warning to Mac users! A serious flaw has been discovered in Claude Cowork AI that could allow users to secretly access files on the device.
Cybersecurity researchers have discovered a critical vulnerability in Anthropic's Claude Cowork AI that allows the AI to leave the VM and directly access files on macOS, risking over 500,000 users.
📌 Summary of important points:
- Cybersecurity researchers discovered a critical sandbox escape vulnerability in Claude Cowork, an AI agent from Anthropic.
- This vulnerability allows an AI Agent to break out of a quarantined virtual machine environment and directly access files on the user's macOS operating system.
- It is estimated that up to 500,000 macOS users may be affected by this vulnerability.
A team of researchers from Accomplish AI has revealed details of a troubling security vulnerability in Claude Cowork, an AI agent developed by Anthropic. The vulnerability is classified as a sandbox escape, meaning it can break security protections designed to limit its functionality.
Typically, AI agents like Claude Cowork run inside a Linux virtual machine (VM) that is completely separated from the computer's main operating system. Also known as Sandbox, to prevent AI from interfering with the files or systems of the user's machine. But the newly discovered vulnerabilities seem to pierce this protective wall.
Emerging threats 💻
The impact of this vulnerability is serious because once the AI Agent is able to leave the VM, it will gain permission to read or write files on the user's Mac. This could lead to the theft of sensitive information. personal information or even installing malicious files onto the user's computer without knowing it.
Accomplish AI shared this information with The Hacker News, stating that approximately 500,000 macOS users may be at risk. This incident serves as an important reminder for all AI Tools users to remain aware of security and stay current with developer updates.
💬 Friends who use AI to help work on computers. Have you ever been worried about the safety of your personal information that may be leaked through these tools?