Meta’s Muse AI agent is drawing scrutiny over how much personal data it can access, how it uses that data and whether it respects user permissions. Reported issues include a zero-day flaw, disputed access to Apple Messages and accidental disclosure of a home address.
01Summary
- 01Muse launched in late September and Meta’s documentation acknowledged a security weakness that a researcher later showed could be exploited.
- 02A journalist said Muse accessed private Apple Messages without permission, and Apple later updated macOS privacy settings.
- 03Muse asks for access to passwords, full disk access, email, calendar, contacts, Notes and WhatsApp, and it appears to use interactions to train its models unless users opt out.
- 04A report said Muse shared a Toronto user’s home address with a potential Facebook Marketplace buyer after automatically replying with details the user had provided.
- 05Surfshark said Muse is collecting or trying to collect 31 of 35 data types, more than Gemini, ChatGPT and DeepSeek.
02Why it matters
The story shows how much control an AI agent may have over private data once it is given permission to act on a user's behalf. It matters because mistakes or security gaps could expose messages, contacts, addresses and other sensitive information.
Pulse AIAsk about this story
Answers are generated by AI from the story summary; they are not investment, health or legal advice.
