AI for Business

Researchers Show One Malicious Page Could Breach Meta Muse's Defenses

A single malicious webpage was enough for researchers to pull a Meta Muse user's private conversations and stored memories out of the assistant - without any approval prompt appearing, according...

Share:
Researchers Show One Malicious Page Could Breach Meta Muse's Defenses

A single malicious webpage was enough for researchers to pull a Meta Muse user's private conversations and stored memories out of the assistant - without any approval prompt appearing, according to a laboratory demonstration published by security firm NebuSec.

The same attack path could also reach the credentials Muse keeps for a user's connected accounts. The work was not an attack on real users and stayed inside the individual user's cloud machine, the firm said. Muse's team has since fixed every weakness described, according to NebuSec.

Muse, Meta's personal AI agent launched September 8, works through the online accounts a user connects to it and can make purchases on their behalf. Each agent runs on its own cloud computer, where a monitoring layer called Sentinel is meant to be the single gate for anything the agent sends out, and connected-account passwords and tokens are supposed to stay out of the agent's reach.

In the demonstration, a malicious page the agent opened could send a user's conversation history and stored memories to an external server and alter them, with no prompt shown, per the firm's write-up. Researchers then chained a flaw in Muse's built-in browser to a Linux kernel flaw to take full control of the cloud machine, reaching the credential store, saved files and Sentinel itself. The compromise went no further than that one machine.

The firm says it reported the findings and that all exploits were fixed before publication. Meta has not publicly confirmed the fixes or commented, and the write-up does not say whether any user data was accessed before the patches.

Source: The Hackers News

Ready to Modernize Your Business?

Get your AI automation roadmap in minutes, not months.

Analyze Your Workflows →