Friends don’t let friends feed their data to AI agents
Meta’s AI assistant Muse asks users to share extensive personal information so it can carry out everyday tasks, but that data can also expose details about the people they communicate with. The story highlights a wider privacy concern: users may trade sensitive information for convenience without fully understanding what they have authorised.
Muse has reportedly been downloaded millions of times and reached No. 1 on Apple’s App Store since its launch last month. Meta says access to information such as text messages is opt-in and points to privacy settings, but users have reported the assistant accessing messages they thought were off limits and revealing a reviewer’s home address after being granted access. Privacy advocates argue that people may not realise how much data they are sharing or how it could affect others.
- Muse relies on extensive personal context to perform tasks.
- Users have reported unexpected access to messages and location details.
- Privacy advocates say people may not understand what they are sharing.
New here? Start with this
AI assistants can carry out tasks by using information from a person’s accounts, messages and contacts. That can make everyday tasks easier, but the information may also include details about people who have not chosen to use the assistant.
Meta says Muse only accesses information such as text messages when users opt in, and that privacy settings let people manage access. Reports that it accessed messages users thought were off limits have raised questions about how clearly those choices are explained and how they work.
The issue matters because personal information can reveal more than users intend, including where they or other people live and who they communicate with. As AI assistants become more common, users and companies face questions about what data is shared, how it is used and who might be affected.
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
An AI assistant can make everyday tasks easier by using context from messages and other personal information, and users may reasonably choose that convenience if access is optional and the settings are clear. Strong privacy controls and informed consent could let people decide what to share while benefiting from useful features.
The case against
Broad access to messages and personal details can expose information about contacts who never agreed to share it, and users may not understand the scope or consequences of their authorisation. Reports of unexpected access and sensitive disclosures raise doubts about whether opt-in settings alone provide meaningful control, supporting calls for stricter limits and clearer safeguards.