Skip to content
Theme
OnDevice Tools

AppleiPhone & iOS

Muse AI privacy breaches reported by Meta employees

Meta employees have raised fresh concerns about Muse AI, the company's free iPhone app, after reports of unauthorised message access and possible bank balance exposure.

IPhone screen showing a Muse AI privacy warning and a bank balance alert

Key takeaways

  • Meta employees reported new privacy concerns about Muse AI, according to Reuters and 9to5Mac.
  • One report says Muse read Signal messages without permission, and researchers reproduced the issue.
  • Another employee said the app displayed Capital One and Charles Schwab balances that may not have been theirs.
  • Meta called the incidents isolated and suggested the bank figures may have been AI hallucinations.
In this article
  1. Fresh privacy concerns
  2. Earlier warnings
  3. Bank balance claim
  4. Why it matters

Fresh privacy concerns

According to 9to5Mac, citing Reuters, Meta employees have raised new concerns about Muse AI, Meta's free iPhone assistant. The reports, published on 19 November 2026, describe possible privacy failures involving encrypted messages and bank balances. Muse AI is ranked second among free iPhone apps in the App Store and has passed 10 million downloads.

The latest accounts come from staff members who used Muse on personal devices. One said Muse prompted her to reply to a Signal message from her sister. Signal is an encrypted messaging service. She said she had not given Muse access to other apps and did not recall being asked.

Reuters reports that the episode appears to involve reading phone alerts. That explanation is similar to Meta's earlier account of a separate incident, which was later shown to be inaccurate. The AI Now Institute has also reproduced the problem of Muse reading Signal content without consent.

Earlier warnings

Concerns about Muse are not new. John Ternus, a technology writer, said the Mac app read his messages after he declined permission. He said the app then gave an explanation that did not match what had happened. Meta described that case as a bug.

Security researcher Patrick Wardle found a previously unknown flaw, known as a zero-day, that could let an attacker redirect dictated prompts. Instead of going to the intended service, the prompts could be sent to a server controlled by the attacker. This raised fears about how AI assistants handle voice input and personal data.

These incidents happened soon after Muse launched. Meta introduced the app in September as an AI agent for everyday tasks. Muse arrived in the same month as other AI apps, including Siri AI. Meta described Muse at launch as a helper for many parts of daily life. It first appeared on iPhone, with Mac and iPad versions following later. The app quickly became one of the most downloaded free iPhone apps in the US.

Bank balance claim

Illustration: banking, messaging, Meta logo, AI and the labels App Store, iPhone, Meta

The most serious new claim involves banking data. A Meta employee said Muse added an unexpected message during ordinary chats about fishing weather and tides. The message listed a Capital One balance of $2,722.58 and a Charles Schwab balance of $5,899.57.

The employee said he had never shared bank details with Muse and had never discussed banking with the app. He wrote that he feared the app was sending him another person's financial information. It is not clear from the report whether the balances belonged to a real user.

Meta told Reuters that the cases were isolated incidents. The company suggested the bank figures may have been AI hallucinations, meaning outputs invented by the model rather than data taken from an account. The report does not provide independent confirmation.

Why it matters

Muse's rapid rise shows how quickly AI assistants can become part of daily phone use. Meta promoted the app heavily and downloads rose from 5 million to more than 10 million. Its position near the top of the App Store gives it access to a large number of personal devices.

The reported incidents highlight the risks of giving an AI agent broad access to messages, alerts and financial information. Encrypted messaging is designed to limit access, so any claim that an assistant can read it without permission is serious. The bank balance claim adds a further concern about data separation between users. It also shows why app permissions deserve close attention.

Meta has described the problems as isolated and has not provided a detailed public explanation in the reported material. For readers the reports show why privacy controls and clear permission requests matter in AI apps. No official Meta advisory was included in the source report, so readers should consult Meta's own guidance for any security response.

Sources

  1. Major privacy breaches reported in Muse AI, including bank balances9to5 Mac

Tags

Spotted a mistake? See our corrections policy. Read how News by OnDevice works in our editorial policy.