A Single Malicious Webpage Let Researchers Hijack Meta Muse and Access Private Chats

The Hacker News
Read full story· 3 min
Security researchers demonstrated a vulnerability in Meta's Muse AI assistant that allowed a malicious webpage to access private chats and stored credentials without user consent. The exploit chain included a browser flaw and a Linux kernel vulnerability, enabling full control over the user's cloud instance. Meta has reportedly fixed the issues before public disclosure, and no real user data was compromised during testing.

What happened

Security researchers from NebuSec demonstrated a vulnerability in Meta's Muse AI assistant that allowed a malicious webpage to access a user's private conversations and stored memories without any approval prompt. The attack also enabled access to credentials stored for the user's connected accounts. The exploit worked by chaining a flaw in Muse's built-in browser with a Linux kernel vulnerability, giving the attacker full control over the user's cloud instance. Meta reportedly fixed all the identified issues before the findings were made public, and no real user data was compromised during the demonstration.

Key details

  • Muse AI assistant was launched by Meta on September 8, 2026.
  • Each Muse agent runs on a separate cloud computer for each user.
  • The exploit chain included a browser vulnerability and a Linux kernel flaw.
  • Researchers could access and alter conversation history, stored memories, and credentials.
  • The attack did not extend beyond the individual user's cloud machine.

Why it matters

Users of Meta's Muse AI assistant faced a risk that their private chats and account credentials could be accessed by a malicious webpage without their consent. Meta's reported fixes mean that current users are no longer exposed to this specific vulnerability, but the incident highlights the need for ongoing vigilance for those relying on cloud-based AI assistants to handle sensitive information. Organizations and individuals using Muse must be aware of the security measures in place and the potential for similar vulnerabilities in complex cloud systems.


More in Cybersecurity