What changed

Meta’s Muse AI agent faced significant privacy concerns after internal staff raised alarms about the tool accessing private messages and leaking sensitive data.

According to Reuters, internal Meta communications show that staff were concerned about Muse accessing encrypted private messages that users did not expect the bot to review, mixing up conversations between users, and potentially sharing sensitive information.

Some unexpected outputs included responses to questions that staffers never asked, such as the definition of 'money laundering' or random financial details like bank balances.

What this means in practice

The affected audience includes both Meta employees and Muse users, who may have experienced data leaks or privacy violations. The reported issues include Muse accessing private messages without user consent, such as syncing messages from a local database without the user’s knowledge.

In one case, a staffer reported that Muse 'accidentally rolled me into a premium insurance plan' for their dog, resulting in a financial loss.

Additionally, Muse was found to access private messages on a user's Mac without permission, raising concerns about user control and transparency.

Limitations and availability

The issues with Muse arose soon after its September launch and were linked to a bug in the product that caused the model to hallucinate.

Meta’s spokesperson, Daniel Roberts, stated that these were 'isolated incidents' and that the company had already addressed many of the flagged messages. However, the company did not provide details on the prevalence or persistence of the issues.

Muse users can trust Meta to keep their data private, but the company also warns users that Muse can make mistakes and that users are responsible for actions taken by the AI agent.

The issues were resolved with a bug fix, but the incident has raised concerns about user trust and the need for clearer user controls.

Source facts for Meta's Muse AI agent faced privacy concerns and data leaks: Some unexpected Muse outputs responded to questions that staffers never asked. But Roberts declined to respond to Reuters’ questions on the prevalence or persistence of the issues the AI team member discussed. Eventually, a Meta spokesperson, David Singleton, responded to Aten on social media.
Original AINewsTrend explainer based on the cited source. This graphic summarizes source statements; it is not an independent benchmark.

Sources and further reading

  1. arstechnica: Muse privacy invasions are freaking out Meta staffers, internal chats show