Urgent.News

What's breaking now, across thousands of outlets.

AI

Meta Muse can keep files on people who have never used it — here's what I found

Meta Muse can build detailed profiles of your friends and family, even if they don't use the AI agent themselves. Here's what you need to know.

Meta Muse can keep files on people who have never used it — here's what I found

Meta Muse, an AI assistant, can potentially retain information about individuals who have never interacted with the service, according to internal instructions extracted by an AI safety and security researcher. The AI can create a dedicated page for every person in a user's life, including family, friends, colleagues, and even people they follow.

These pages are designed to accumulate information over time, covering facts, history, relationship dynamics, shared interests, and methods to strengthen connections. However, the detailed instructions show that Muse is explicitly programmed to rely on available evidence rather than inventing details, which serves as a safeguard against making up information about others.

While Meta assures that each user has a separate virtual machine storing their data and context, the potential for information to be shared across users' Muse agents if they are connected in someone's life raises privacy concerns. Users have the option to remove memories or disconnect external services, and Muse includes an audit log to monitor its activities.

Written by urgent.news from Tom's Guide's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at tomsguide.com →

More in AI

[Editor's Pick] Who Bears the Cost and Risk of AI?

The financial and accountability structure surrounding artificial intelligence is becoming more complex. Wall Street is arranging $60 billion to support Anthropic’s access to AI chips, a U.S.

  • Wall Street pools $60 billion to fund Anthropic's GPU chip access.
  • Reflection AI launches open-weight AI model Beam on October 5.
  • Financial institutions, enterprises, and policymakers share AI costs and risks.

Your Agent Says Verified. The Check Behind It May Be Empty.

Ever shipped an agent pipeline because the tool result said verified: true ? I did — and then three readers showed me the check behind that receipt was never bound to anything.

  • Engineer found issue with agent pipelines reporting verified: true without actual content check.
  • Verifier recorded hash of bytes checked, not running predicate to verify content.
  • Engineer fixed issue by exposing empty verifier using known bad artifact.

More from Tuesday 6 October →