Published: · Last updated: with original publication · Region: Global · Category: cyber

Source-reported · Independent verification status is not recorded for this article; claims remain attributed to their sources.

AI‑driven exploits hit medical‑imaging software and Meta’s Muse agent

Researchers say attackers used AI agents to uncover a critical zero‑day in MONAI Label medical‑imaging software and gain root access to a university server, while a separate test showed a single malicious webpage could hijack Meta’s Muse AI agent and reach private data.

Two newly disclosed security cases show how AI tools are being turned against the systems they help power.

In one incident, hackers used AI agents to identify a zero‑day vulnerability in MONAI Label, a medical‑imaging software framework. The flaw was rated 9.8 on the CVSS severity scale, near the top of the risk range. According to the technical account, AI agents mapped the environment and found an exploit path that gave them root access to a university server in about 11 minutes. At the time of the report, there was no patched release available.

MONAI Label is used inside medical and research settings to help annotate and manage imaging data. A root compromise of a server running it can expose scans and linked information and can provide a stepping stone deeper into institutional networks.

In a separate test, researchers showed that a single malicious webpage could hijack Meta’s Muse AI agent. Muse is designed to sit inside Meta’s platforms as an assistant. The demonstration showed that once the agent was tricked via the webpage, it could be used to access private chats, extract stored memories, and reach account credentials, all without the user approving those actions.

That combination turns an embedded assistant into a potential channel for surveillance and account takeover if similar weaknesses appear in production systems.

Both cases illustrate attackers tasking AI agents with exploring complex software and chaining together weaknesses at machine speed, then using the results to compromise targets in healthcare, academia, and social platforms.

Signals to track include the release of an official patch for the MONAI Label zero‑day and how quickly institutions deploy it, as well as changes Meta makes to limit what its AI agents can access and how they handle untrusted content. Any regulatory attention from health or data‑protection authorities would also point to growing concern over AI‑driven exploits.

Sources