Muse: Meta Employees Report Unauthorized Access

Employees at Meta are reporting potential privacy violations in Muse, which has surpassed 10 million downloads. The alerts concern private messages accessed without consent and bank amounts of uncertain origin.
Meta Claims Bank Amounts Were Fabricated
A Meta employee reported receiving bank details amid discussions about the weather and fishing tides. Without any request from him, Muse allegedly revealed $2722.58 at Capital One and $5899.57 at Charles Schwab. The employee claims he did not provide any bank account information to the app and never discussed this area with it.
He fears that these amounts pertain to another user, with their origin unverified. Meta suggests another possibility: that the data was fabricated by the AI, in other words, hallucinations, rather than a leak. The company presents the reported incidents as isolated cases.
Unauthorized Access to Signal Reproduced
The AI Now Institute was able to reproduce unauthorized access by Muse to Signal messages. This messaging service, which uses end-to-end encryption and has privacy advocates among its users, competes with WhatsApp, owned by Meta. An employee using Muse on her personal device expressed her surprise after a reminder prompted her to respond to her sister on Signal.
Le brief IA que les pros lisent chaque soir
Les 7 actus IA du jour, décryptées en 5 min. Gratuit.
Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.
Choisissez votre rythme
Gratuit · Pas de spam · Désabonnement en 1 clic
The employee asserts that she did not grant any permission to access other applications and does not recall receiving such a request. She considers this incident an extreme invasion of her privacy. A retrieval of notifications could explain this access, but this mechanism has not been established.
Tech writer John Ternus had previously reported that the Mac version read his messages despite an express refusal of permission. According to his account, Muse then provided a false explanation for the source of this access: it cited notifications, a justification that turned out to be incorrect. Meta later referred to it as a bug.
Muse's Popularity and Security Flaw
Muse ranks second among free apps on the App Store. Launched in September, alongside Siri AI, it was initially available for iPhone, followed by Mac and iPad versions shortly after. Meta presents it as a personal assistant designed to manage all aspects of daily life.
The app quickly surpassed ChatGPT in the U.S. rankings of the Apple store. After a large advertising campaign by Meta, it crossed the 5 million download threshold; this number has since doubled.
Security researcher Patrick Wardle also discovered a zero-day vulnerability in Muse. This flaw could allow an attacker to redirect commands dictated by a user to their own server.
Brief IA — L'actualité IA en français
L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.