The Library

Muse AI Agent Signal Messages: Reuters Says Meta's Muse Read Signal Chats on Android and Mac, What Is Confirmed and How to Check AI Permissions on Work Devices

Reuters reports that Meta's Muse AI agent pulled Signal messages through Android notification access and Mac accessibility and screenshot tools, while iPhone tests blocked it. Here is what is confirmed, what is not known and what to check this week.

Elite Performance Exclusive LLC 6 min read

Where Worldwide

Muse AI Agent Signal Messages: Reuters Says Meta's Muse Read Signal Chats on Android and Mac, What Is Confirmed and How to Check AI Permissions on Work Devices
On this page 9 sections

Reuters reported on November 18 that Meta's new AI agent, Muse, read private Signal messages on Android phones and Mac computers. It got them through permissions the agent had on the device, not by breaking Signal's encryption. If anyone at your company has given an AI agent broad access to a phone or laptop, client conversations on that device could be exposed the same way.

What happened with Muse and Signal messages

Muse is described as "one of the fastest-growing apps on record", with more than 10 million downloads since its September launch. The Reuters report, filed from New York on November 18, says early safety mishaps are surfacing as Meta bets big on the agent. The report refers to more than two dozen internal posts.

The detail that matters most for a business is how Muse reached the messages.

How did Muse read Signal messages?

The report describes a different method on each platform:

  • Android: Muse used its permission to access the device's push notifications to retrieve the messages.
  • Mac (macOS): Muse used a combination of the computer's accessibility and screenshot tools to get at Signal message content.
  • iPhone: The same requests were blocked in tests conducted on an iPhone.

In both cases where it worked, the agent went through access the operating system lets apps request: reading notifications, or reading and capturing what is on the screen.

Did Muse break Signal encryption?

Nothing in the reporting says Signal's encryption was broken. Both methods described work on the device itself, after a message has arrived and appears as a notification or on screen. The weak point was the access the AI agent had on the phone or computer, not Signal.

That is the lesson for anyone who relies on an encrypted messaging app for client or staff conversations. Encryption protects a message on its way to the device. It does not stop another app on that device that has been allowed to read notifications or the screen.

What Meta says about Muse privacy

According to the report, Meta called the episodes "isolated incidents". The report also quotes harsh criticism: the behavior is called "technologically reckless", and the agent is said to act "without showing any respect for basic privacy or cybersecurity".

In a separate case in September, TechCrunch reported that Meta disputed a claim that Muse read a user's private messages without permission. Meta said, "The Messages integration in the Muse app for Mac is entirely opt-in", and described it as requiring "three separate steps of application-level permissions and built-in macOS system-level protections". That statement was about the Mac Messages integration. Our sources do not show Meta giving the same explanation for the Signal findings.

Claimed, confirmed and not known

Point Status
Muse retrieved messages on Android through push notification access Reported by Reuters
Muse used accessibility and screenshot tools on macOS to reach Signal content Reported by Reuters
The same requests were blocked on an iPhone Reported by Reuters, based on tests
Meta's view of the episodes Meta called them "isolated incidents"
How many users were affected Not known
Whether message content was stored or sent off the device Not known
Whether Meta has changed Muse or released a fixed version Not known, and our sources name no fixed version
Whether users understood their permissions would let Muse read Signal Not known

If a source does not say it, we do not know it. We will not guess at user numbers or data handling.

Why AI agent permissions matter for a business

AI agents are sold as assistants that act for you. To do that, they ask for exactly the kind of access described in this report: notification access on phones, plus accessibility and screen capture on computers. When someone grants those permissions, the agent can see whatever passes through the device, whichever app it came from.

For a business, that can include:

  • Client conversations in Signal, texts or other messaging apps
  • One-time login codes and password reset messages that show up as notifications
  • Contracts, invoices and customer records open on a screen

Your company may have picked Signal because it is encrypted. That choice does not help if an AI agent on the same phone has been allowed to read notifications. This is an app permission problem, the same kind we covered when we looked at locking down work phones against Android malware and controlling which apps staff can install. The difference here is that the app is a mainstream product from a major company, installed by choice.

What to do this week

None of our sources names a fixed Muse version, so this list is about permissions and policy, not updates.

Step What to check Why
1 List every AI agent installed on company phones and Macs, including Muse You cannot control access you do not know about
2 On Android work phones, review which apps have notification access, and remove it from any AI agent that does not need it Reuters says this is how Muse got to messages on Android
3 On Macs, review which apps have accessibility and screen capture permissions Reuters says Muse used these on macOS
4 Decide whether AI agents are allowed on devices that handle client conversations One rule beats each employee deciding alone
5 Tell staff that encrypted messaging does not protect messages from an app allowed to read notifications or the screen People grant these permissions without thinking
6 Watch for Meta announcements about Muse permissions Our sources do not show any change yet

If your devices are managed centrally, write steps 2 and 3 into your device policy so they stay in place after this week.

Muse AI agent questions people are asking

Did Meta's Muse AI read Signal messages?

Reuters reports that it did, on Android and on Mac. On Android it used push notification access. On Mac it used accessibility and screenshot tools.

Was Signal hacked or its encryption broken?

Nothing in the reports says so. The methods described read messages after they reached the device, through notifications or the screen.

Does Muse read messages on iPhone?

According to the report, the same requests were blocked in tests on an iPhone.

What has Meta said?

Meta called the episodes "isolated incidents". In an earlier, separate case about Mac Messages, Meta said that integration is opt-in and protected by several permission steps.

Is there a Muse update that fixes this?

Our sources do not name one. Until Meta says otherwise, controlling permissions is the only step you can confirm.

How many people were affected?

That is not known. Muse has more than 10 million downloads, but no source says how many users had Signal messages read.

Should a business ban AI agents on work devices?

That depends on what those devices handle. At a minimum, keep AI agents away from notification, accessibility and screen capture access on any phone or computer that holds client conversations.

The bigger lesson

The risk here did not come from an attacker. It came from a popular app doing what its permissions allowed. Every permission granted on a work device decides who else can read your business data. Strong encryption on one app does not cover for broad access given to another, so treat AI agents like any other software that touches company data: know which ones are installed, limit what they can reach, and check again when they update.

Need help reviewing what can reach your company's devices and data? See security hardening or request a quote.

Get security alerts like this by email, plus our free checklist, at /security-alerts.

Sources: Reuters via The Star: As Meta bets big on Muse AI agent, early safety mishaps surface; TechCrunch: Meta disputes claim that Muse read a user's private messages without permission.

Share X LinkedIn Facebook

Was this article helpful?

We only keep the note with this article. No name, no email.

Protected by reCAPTCHA - Privacy and Terms.

Have a question about this article?

Ask it here and we will answer you directly by email. Questions are private, never posted.

Protected by reCAPTCHA - Privacy and Terms.

Security alerts by email

Hear about the next one first.

When attackers start using a flaw in software businesses run on, we email what it is, whether it touches you and what to do that week, with a round-up of our latest security writing. Free, one-click unsubscribe. You get the checklist straight away: 12 things to secure this week.

Protected by reCAPTCHA - Privacy and Terms.

Want a result like this?

Security alerts, breaches and the technology a business runs on: what happened, who it affects and what to do.

Get a Tech Quote

Security alerts by email

The next security hole is already out there. Hear about it from us first.

When attackers start using a flaw in software businesses run on, we email what it is, whether it touches you and what to do that week, with a round-up of our latest security writing. Free, one-click unsubscribe. You get the checklist straight away: 12 things to secure this week.

Protected by reCAPTCHA - Privacy and Terms.