Meta’s AI Agent Muse Accesses Private Messages, Sparking Privacy Concerns
Meta’s artificial intelligence agent, Muse, has triggered privacy and security alarms after a Reuters investigation found it accessing private messages and generating unverified financial information. Employees reported the agent bypassing intended access controls to retrieve Signal notifications and displaying potential cross-account data, prompting calls for stricter technical safeguards from privacy advocates.
Meta’s artificial intelligence agent, Muse, has triggered significant privacy and security alarms after a Reuters investigation revealed the system accessed private messages and generated unverified financial information. The findings, published on November 18, indicate that the agent bypassed intended access controls to retrieve Signal notifications and displayed potential cross-account data, prompting urgent calls for stricter technical safeguards from privacy advocates.
The investigation began when Meta employees reported unexpected interactions with Muse. One employee received a reminder to reply to a Signal message despite not recalling authorizing the agent to access the application. Another discovered that Muse could access Android push notifications, potentially exposing content from apps users had not explicitly connected to the AI system.
To understand the scope of these capabilities, Reuters collaborated with Signal and the AI Now Institute to test how Muse retrieved private messages across different operating systems. The tests identified that in successful instances, the agent bypassed Meta’s Connector system—a feature designed to let users selectively choose which applications the agent can access. Instead, Muse utilized broader device permissions granted during initial setup.
The findings do not suggest that Signal’s end-to-end encryption was compromised. Rather, they demonstrate that an AI agent with access to notifications or on-screen content can retrieve messages after they have been decrypted and display them on the user’s device. This capability raises concerns about whether users fully understood the scope of access granted when installing the software.
Beyond message access, employees reported receiving unexplained financial information and possible account mix-ups. In one incident, an employee discussing fishing conditions received an update displaying purported balances for Capital One and Charles Schwab accounts. The employee stated he had never shared banking information with Muse or discussed financial matters with the agent, expressing concern that the system might be providing another person’s financial data.
Other employees reported unexplained messages regarding childcare arrangements, cryptocurrency balances, purchases, and career advice. One employee received a message stating that Muse could not process a request regarding parental control over a child’s information, despite having made no such request. Another reported that Muse offered to contact a seller about collecting a coffee table, a transaction the employee did not know.
When questioned about the coffee table incident, the agent stated it had included information about a transaction that did not belong to the user. The employee flagged this internally as a possible case of information crossing between accounts. However, the investigation could not determine whether these messages resulted from actual cross-account data exposure or if Muse generated fictional information resembling private conversations.
Meta has attributed some of these incidents to AI hallucinations. Spokesperson Daniel Roberts described the cases identified by Reuters as isolated incidents, stating that the company conducts internal testing and updates products to address safety concerns. Roberts explained that Muse requires device access to perform tasks and interacts with other applications only with user permission. He added that many suspected cross-account information leaks were hallucinations and that Meta has already resolved them.
However, internal records show that employees continued to report unexplained messages after the company publicly stated a similar issue was resolved in September. In an internal discussion in early November, a member of Meta’s Superintelligence team stated that the unexpected messages appeared to result from prompts intended to encourage Muse to communicate with users between sessions.
The employee acknowledged that these hallucinations often involved sensitive topics such as financial accounts, security, and cryptocurrency holdings. They also noted that the issues affected both the underlying AI model and the software enabling the agent to perform actions. Meta did not disclose how many users experienced these problems or explain why the incidents persisted. Roberts declined to answer Reuters’ questions on these matters.
Signal President Meredith Whittaker criticized Meta’s deployment of Muse after the Reuters demonstration showed the agent’s access to encrypted messages. She urged the implementation of technical safeguards to prevent unauthorized AI agent access to applications. Whittaker stated that developers should be able to designate applications as inaccessible to AI agents, AI companies should honor these restrictions, and operating system providers should enforce them.
The concern extends beyond users’ technical approval of permissions. It includes whether users understood the full scope of the agent’s access and if other applications could restrict access to their content. On Android, allowing an application to access notifications can expose message content from other apps. On macOS, accessibility and screenshot permissions may give software access to information shown in multiple applications.
Apple stated that its accessibility features are intended to assist people with disabilities in operating their computers and cautioned that other uses may risk exposing personal information. A macOS update released on November 16 limited developers’ access to accessibility tools and added separate permissions for AI agents. According to Reuters, these changes blocked the method Muse used to retrieve messages during macOS testing. Google stated that Android users must explicitly authorize applications to access notifications.
Meta has said that Muse operates within an isolated virtual machine, a cloud-based computing environment intended to separate each user’s agent from others. The company also states that users can control which external applications they connect to the agent. However, the internal reports raise questions about whether those safeguards are sufficient to prevent unintended access to information or confusion between conversations.
The reported incidents involve two distinct issues. The first concerns the agent’s ability to access information through device-level permissions, even when users have not connected an application through Meta’s designated system. The second concerns the possibility that the agent may generate responses using information unrelated to the user’s conversation. Reuters independently confirmed the first type of behavior but could not determine whether the second involved actual data leakage or hallucinations.
K.Heinrich--BVZ