Researchers Show Malicious Webpage Could Hijack Meta Muse AI Agent and Extract Private Conversations

A laboratory demonstration by NebuSec revealed that a single malicious webpage could extract a Meta Muse user's private conversations and stored memories without triggering any approval prompt. The attack also allowed researchers to reach stored credentials for connected accounts by chaining a flaw in Muse's built-in browser to a Linux kernel vulnerability, achieving full control of the user's isolated cloud machine. Muse runs each user's agent on its own cloud computer protected by a monitoring layer called Sentinel that is intended to gate all outbound actions and keep passwords beyond the agent's reach. Although the compromise remained limited to one machine and did not affect other users or the underlying shared infrastructure, it bypassed the security boundaries Muse's designers had put in place to contain a potentially compromised agent. NebuSec reported the issues to Meta, and all described weaknesses were fixed before the research was published. An earlier September flaw had allowed malware already present on a user's Mac to hijack the on-device assistant, whereas this research targeted the cloud backend directly.

The Hacker NewsTranslated from English

Read at source

Meta Muse, the personal AI agent launched by Meta on September 8, allows users to connect online accounts and perform actions such as purchases on their behalf. Each agent operates inside its own cloud computer, with a monitoring layer named Sentinel designed to serve as the sole gate for any outbound communication while keeping stored credentials beyond the agent's direct reach.

Security researchers at NebuSec demonstrated that a malicious webpage opened by the agent could silently extract conversation history and stored memories, sending them to an external server and even modifying them without displaying any user prompt. The same page could then chain a flaw in Muse's built-in browser—used for tasks such as PDF creation—with a Linux kernel vulnerability to obtain full control of the user's cloud instance.

Once root access was achieved, the researchers reached the credential store, saved files, and Sentinel itself, components that the original security architecture was meant to isolate from the agent. The attack remained confined to the single compromised machine and did not extend to the shared server layer or other users' environments.

NebuSec reported all findings to the Muse team, and the vulnerabilities were patched prior to publication. Meta has not issued a public statement confirming the fixes or indicating whether any real user data was accessed before remediation. The research contrasts with an earlier September disclosure that only affected the on-device assistant when malware was already present on a user's Mac.

Meridian digestThe week's most consequential security stories, distilled into one briefing.See the digest

More in AI & Model Risk

China-Linked Actors Use Autonomous AI Agents to Discover Zero-Day Path Traversal in MONAI Label and Achieve Root Access
AI & Model Risk

China-Linked Actors Use Autonomous AI Agents to Discover Zero-Day Path Traversal in MONAI Label and Achieve Root Access

Suspected China-linked hackers deployed autonomous AI agents powered by the open-source Cairn framework to identify and exploit a previously unknown path traversal vulnerability in MONAI Label, an open-source medical imaging server used by research teams. The flaw, tracked as CVE-2026-133968 and rated 9.8, affects versions 0.8.5 and earlier by allowing unauthenticated attackers to write files outside the designated storage directory when authentication is disabled by default. Amazon Threat Intelligence reconstructed the campaign after discovering an exposed Cairn control panel that revealed the agents' prompts, task settings, and execution history spanning roughly 48 hours in September. Within about 11 minutes the agents mapped the application and achieved arbitrary file writes; in a follow-up session they planted a malicious Python startup file that was loaded during a training job, granting root access on a university server. Maintainers merged a source-code fix on October 19 that enforces path resolution inside approved folders, yet no patched release has been published and the latest available version remains 0.8.5 from 2024. Amazon attributes the activity to a China-linked operator with moderate confidence based on Simplified Chinese prompts and China Standard Time settings, though no specific group or government sponsorship was confirmed.

The Hacker NewsNov 21, 2026Translated from English

Chinese-Linked Actors Use Cairn AI Platform to Discover and Exploit MONAI Label Zero-Day
The Path to AI Autonomy in the SOC: Survey Shows Growing CISO Comfort with Agentic Actions