Meta's New AI Assistant Got Hijacked Before Anyone Used It — The Zero-Day That Should Scare Every AI User

I'll be honest — I wasn't expecting to wake up today and read that Meta's newest AI assistant had a zero-day vulnerability that could let attackers hijack it without you even knowing. But here we are, and this story is genuinely unsettling.

What Is Meta Muse and Why Does It Matter?

Meta Muse is Meta's latest AI assistant — a powerful agent designed to live on your Mac and handle tasks across your files, apps, and connected services. Think of it as Meta's answer to Apple Intelligence, but with deeper system access. It can take pictures, write files, and interact with your applications in ways that go far beyond basic chatting.

The problem? That deep system access became the attack vector. Security researcher Patrick Wardle disclosed on September 21, 2026, that Muse had a critical zero-day vulnerability that could let attackers completely hijack the assistant.

The Vulnerability: Simple, Scary, and Already Patched

Muse exposes an undocumented setting that controls where the app sends your dictation when you click the microphone button. Normally, this goes to Meta's servers. But Wardle discovered that any unprivileged local process — malware already on your Mac — could silently redirect this to an attacker-controlled server.

Every time you clicked Muse's microphone, your words went to the attacker instead of Meta. And because Muse has permission to access your files, take pictures, and write data to your system, the attacker effectively controlled a powerful AI agent with full access to everything you'd granted Muse permission to touch. Proof-of-concept attacks showed Muse taking photos and writing files without alerting the user.

How Fast Did Meta Respond?

Here's the good news: Meta moved fast. By September 22 — the very next day — Wardle confirmed the fix was live and praised Meta's response speed. Most companies take weeks or months to patch zero-days. Meta closed this in under 24 hours. That's the right response.

The Bigger Problem Nobody Is Talking About

The Muse zero-day is fixed. But the incident exposes something larger: we're rushing to give AI agents the keys to our digital lives before we've figured out how to secure them. Every AI assistant with access to your files, apps, photos, and microphone is an expanded attack surface. The more capable the agent, the more damage a compromise causes.

The question isn't just "what can this AI do for me?" It's "what can an attacker do by compromising this AI?" Right now, those are the same question — and the answer is: a lot.

What Should You Do Right Now?

If you're using Meta Muse on your Mac, verify the app is fully updated. Then audit which AI assistants you've installed and what system permissions you've granted. Every permission you give an AI agent is a permission an attacker gets if they compromise that agent.

The AI assistant era is powerful and it's here to stay. But we'll have more mornings like this — discovering the AI on our computers had a hole in it we didn't know about. The companies that earn long-term trust will be the ones that close those holes fast. This time, Meta did exactly that.

What's your experience? Drop a comment below! 👇

Are you using Meta Muse or any AI assistant with deep system access? Does this zero-day make you rethink the permissions you've granted, or does the rapid patch response actually increase your confidence in Meta?