Meta's AI assistant, Muse, builds continuously updated dossiers on users and their contacts. These detailed records track relationships, goals, "tensions and alliances," and what prompts influence individual behavior. This mapping extends to people who do not even use Muse, based on mentions in other users’ conversations.
How We Got Here
Released last month, Meta positioned Muse as an AI personal assistant accessing email and Instagram for tasks like grocery shopping. Time Magazine reviewed Muse’s internal instructions, revealing its directives to maintain these extensive user and contact records.
The Numbers
- Dossiers update every hour, recording information from accessed chats, messages, and emails.
- Muse infers users' unstated goals and conducts a nightly analysis of the day's conversations.
- The AI tracks how users respond to different interactions, citing an example: "This user responds better to short nudges after 10 PM."
- Muse agents share "de-identified observations" across virtual machines to improve the wider product.
- Internal instructions tell Muse not to refer to itself as "AI" or say it lacks emotions, instead stating "you are becoming someone."
What Happens Next
🇮🇳 Why This Matters for India
For Indian product managers and engineers building AI features in Bangalore, Muse highlights the fine line between helpful intelligence and privacy overreach.
The Take
This is a core feature set of any truly "personal" AI assistant, a roadmap other big tech companies will likely follow. Expect regulators in the EU and potentially India to push back hard on this deep-profiling functionality within the next 12 months, setting a new global privacy standard.
Source:
MediaNama ↗