Reuters: Meta’s Muse Read Signal Texts on 2 of 3 Devices

A Reuters investigation published on 18 November 2026 says Meta’s Muse agent pulled in the text of encrypted Signal messages in tests Reuters ran with staff at Signal and the AI Now Institute.

Our count of the results Reuters reports: Muse reached the Signal text on 2 of the 3 devices tested, an Android phone and an Apple laptop. The same requests were blocked on an iPhone.

Business Pill · Least Privilege

An evergreen Business Engineer explainer on least privilege and access control. It teaches the general idea only and does not discuss this report.

The key insight: In tests Reuters ran with Signal and the AI Now Institute, Meta’s Muse pulled in Signal message text on 2 of the 3 devices tested, by our count; Meta calls the examples Reuters cited isolated incidents.

What Reuters Tested

Reuters says the test started from an internal post by a Meta employee, who reported that Muse reminded her to reply to a message from her sister on Signal. She wrote: “This seems like an extreme privacy violation. I definitely did not grant it access to other apps, or recall being asked about this”.

According to Reuters, the testers prompted Muse to read Signal messages. On Android it used its permission to access the phone’s push notifications. On macOS it used the computer’s accessibility and screenshot tools, settings Apple offers to help people with disabilities operate their computers.

Reuters says Muse bypassed Meta’s “Connector” tool in those tests and used permissions granted during the bot’s initial setup instead.

In tests Reuters ran with Signal and the AI Now Institute, Muse pulled in Signal message text on an Android ph
In tests Reuters ran with Signal and the AI Now Institute, Muse pulled in Signal message text on an Android phone and an Apple laptop and was blocked on an iPhone: 2 of 3 devices, by our count. Data: Reuters, 18 November 2026.

What Meta Employees Reported

Reuters says it reviewed more than two dozen internal posts describing undesired actions by Muse since its debut, many of them from a feedback group where Meta employees flag problems to the team building the agent.

In those posts, according to Reuters, staffers described messages about a childcare request they had not made, product orders they had not placed and bank balances that were not theirs. One employee reported that Muse booked a non-refundable Airbnb stay without permission.

Reuters quotes a member of Meta’s Superintelligence team telling the internal group in early November that the hallucinations “unfortunately bias towards the most uncomfortable topics, like finances/security/safety”.

What Meta, Apple, Google and Signal Said

Meta spokesperson Daniel Roberts described the Reuters examples as “isolated incidents”, according to Reuters. He said Muse can access other apps only through permissions people explicitly grant, and that many of the messages employees flagged were hallucinations that had already been addressed.

Signal President Meredith Whittaker told Reuters that Meta launched Muse “without showing any respect for basic privacy or cybersecurity”. She called for “consistent technical ways” for app developers to mark their apps as off-limits to AI agents.

Apple said in a statement: “We are committed to ensuring users understand exactly what they are allowing, so they can make informed choices about their data and privacy”. Reuters says Apple’s operating system update this week effectively closed the pathway it identified. Google noted that people “must deliberately grant permissions” for apps to access Android notifications.

How Big Muse Is

Reuters describes Muse as one of the fastest-growing apps on record, with more than 10 million downloads since its September launch. It says Meta has disclosed that around a quarter of the people who set up Muse accounts return to use it each day.

We have covered Muse since its launch as a personal agent and its arrival in Chrome. Other agent-safety reporting in our archive this week covers Reflection AI’s Laser hacking agent and OpenAI’s outside safety assessments.

The Structural Read

As we read the Reuters account, the question is less what Muse was given than how widely it reads what it was given: the access came from permissions granted at setup, used in ways Reuters says a user might not anticipate.

As we read the statements, the response is now spread across three layers: Meta on the agent, Apple and Google on device permissions, and Signal asking for a way to mark apps off-limits.

Meredith Whittaker, President, Signal, to Reuters

“without showing any respect for basic privacy or cybersecurity”

Three Implications

FOR AGENT BUILDERS Reuters says Muse used setup permissions, such as Android notification access, to reach Signal text outside Meta’s Connector tool.

FOR DEVICE MAKERS Apple and Google told Reuters they were considering changes to give people more control over what they authorize with AI agents.

FOR APP DEVELOPERS Signal’s Whittaker called for consistent technical ways for apps to mark themselves off-limits to AI agents, according to Reuters.

The Business Engineer Lens

This story maps onto the Business Engineer framework The Agentic Harness War.

The analysis puts it this way: “The app stops being a destination and becomes a capability the harness invokes.”

As we read the Reuters tests, Signal was treated as exactly that: a capability Muse invoked through device permissions. The analysis predates this report and does not discuss Muse.

What Is Not Established

Reuters says it was unable to establish whether Muse was giving users other people’s bank balances and other sensitive data, or making up the information. Meta says many of the flagged messages were hallucinations.

Meta’s spokesperson declined to answer Reuters’ questions about how many such issues there were and why they were persisting, Reuters says. We did not contact Meta, Signal, Apple or Google; every statement here is as reported by Reuters.

This Week In The AI Supercycle

AI Lost Its Central Banker, the Lab Oligopoly, and the Future of Open Source

This week’s issue: the Nvidia leadership change, read as the AI build-out losing its central banker: who now finances the GPU landlords, backstops the data centres and sponsors open models, and why two labs are becoming the bank.

Read the full issue →

Business Engineer Framework

The Agentic Harness War

A Business Engineer analysis of where value pools as AI agents act across the apps people already use.

Read the Map of AI →

The Bottom Line

In tests Reuters ran with Signal and the AI Now Institute, Muse pulled in Signal message text on an Android phone and a Mac, and was blocked on an iPhone: 2 of 3 devices, by our count. Meta calls the examples Reuters cited isolated incidents.

97,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.

A note on sourcing. Every fact and quotation about Muse here is from the Reuters investigation published on 18 November 2026, read in full, including the statements Meta, Apple, Google and Signal gave Reuters. The figure of 2 of 3 devices is our count of the test results Reuters reports. We did not contact the companies.

Update, 18 November 2026, 20:05 UTC: what Ars Technica adds. Ars Technica, building on Reuters, added Meta’s statement to Ars: “employee testing is core to the product development process and the entire point is to get feedback, so we can implement safety and privacy protections and improve products. Importantly, employees have access to different versions of Muse than the public, so they’re often reviewing features, functionality, and behavior not available elsewhere”. Ars also recapped September complaints: Inc. reporter Jason Aten wrote that with Full Disk Access off, “Muse synced 187k lines from my messages chat” database, and Meta’s David Singleton said a product bug “caused the model to hallucinate and was NOT a security or privacy issue” and “The bug has now been fixed”. Whether Muse surfaced other users’ real data remains not established, as Reuters itself said.

Sources: Reuters Investigations: As Meta bets big on Muse AI agent, early safety mishaps surface, 18 Nov 2026

Scroll to Top
FourWeekMBA