
Loading

Loading
We use strictly necessary cookies to run this site, and analytics cookies to understand how it's used. See our Privacy Policy for details.

Meta’s AI agent Muse has raised privacy concerns among employees after it accessed messages from encrypted apps, generated information about transactions users said they had not made and sent chats that appeared to relate to other people, according to a Reuters report based on internal company posts.
Summary
The report noted Muse, which has recorded more than 10 million downloads since its September launch, prompted employees to flag concerns that it could be mixing up conversations between accounts and exposing users’ private information. In one instance, an employee reported that Muse referred to messages received on Signal, an end-to-end encrypted messaging app, despite the employee not recalling granting the agent access to the service. Other employees reported receiving unsolicited messages about bank balances, childcare requests, career advice and product orders that they said were unrelated to their conversations.
Reuters reviewed internal posts documenting such incidents. It also worked with Signal staff and the AI Now Institute to test Muse's access to Signal messages. The tests found that the agent could retrieve message content through permissions granted during its initial setup.
On Android devices, the agent used access to push notifications to retrieve messages. On Apple computers, it used accessibility and screenshot tools. The same requests were blocked during tests on iPhones.
Meta spokesperson Daniel Roberts described the incidents as “isolated incidents,” saying the company prioritises safety, conducts internal testing and updates its products to improve protections. He said Muse can access other apps only through permissions users explicitly grant. He also attributed several messages flagged as possible account mix-ups to AI hallucinations, adding that the issues had been addressed.
The report could not establish whether the bank balances and other sensitive information shared by Muse belonged to other users or were generated by the AI. Signal President Meredith Whittaker criticised Meta's approach, saying the company had launched Muse “without showing any respect for basic privacy or cybersecurity.” She called for app developers to have technical mechanisms to prevent AI agents from accessing their services and for operating system providers to enforce those restrictions. Apple and Google noted they were considering changes to give users more control over the permissions granted to AI agents.
Apple, as per the report, said, its accessibility tools were designed for people with disabilities and that using them “for unintended purposes” could put personal information at risk. The company also said it had updated its operating system to restrict developers' use of those tools and introduce new permissions for AI agents. The incidents come as technology companies develop AI agents that can interact with apps and devices to carry out tasks such as answering emails, booking appointments and making purchases.
The agents require access to users' devices and data to perform these tasks, but their actions can also create privacy and security risks. Meta has said Muse operates within a secure virtual machine and allows users to control which apps they connect. The company also warns users that the agent may make mistakes, while its terms of service make users responsible for its actions.
The report noted that internal posts showed concerns about unexpected messages continuing from September through November.
KazaSec's take
AI-related security incidents are a genuinely new category, prompt injection, model manipulation, and data leakage through an LLM integration don't map cleanly onto traditional application security testing, and are worth assessing deliberately rather than assuming existing controls already cover them.
Coverage details
We've archived 1209 other articles touching the same topic (news, latest stories, meta) , see the full security news archive.
Related security advisories
Relevant from KazaSec
More coverage on this topic
We help organizations find and fix the gaps before they make headlines.