Episode Details

Back to Episodes
Meta Muse: When Your AI Assistant Becomes the Attack Surface

Meta Muse: When Your AI Assistant Becomes the Attack Surface

Episode 581 Published 1 month, 1 week ago
Description

Meta Muse is designed to work across a user’s inbox, calendar, browser, connected apps, and other personal services. That makes it useful—and makes its delegated authority a security problem worth examining.

Tom Eston and Scott Wright discuss Meta’s stated guardrails for Muse, including isolated execution, connector separation, credential boundaries, and approval flows. They also examine reporting on a Muse authentication-token issue and why a compromised token can matter more when an agent is authorized to act across a person’s digital life.

The conversation also covers privacy expectations, reports of human involvement behind AI tasks, liability when an agent makes a harmful decision, and how digital-legacy planning changes when an agent could continue acting after its owner is inactive. The practical takeaway: start with least privilege, separate read from write and spending authority, approve consequential actions, review logs, and keep a fast way to revoke access.

** Links mentioned on the show **

Ars Technica — Muse, Meta’s extraordinarily privileged AI assistant, has a serious 0-day https://arstechnica.com/security/2026/09/muse-metas-extraordinarily-privileged-ai-assistant-has-a-serious-0-day/

Meta AI Research — How We Built Safety Into Muse https://research.meta.ai/blog/security-and-safety-for-ai-agents-our-approach-with-muse

Meta Help Center — How Muse handles your privacy, safety and security https://www.meta.com/help/artificial-intelligence/1047255454427887/

CNBC — Meta pushes into personal AI agents in Muse Spark family https://www.cnbc.com/2026/09/08/meta-personal-ai-agents-public-reckoning-privacy-safety.html

WIRED — Meta Releases Muse, a Personal AI Agent With Privacy Built Into It https://www.wired.com/story/meta-releases-muse-a-personal-ai-agent-with-privacy-built-into-it/

The Guardian — Meta’s AI agent Muse gives out user’s home address without permission https://www.theguardian.com/technology/2026/sep/28/metas-ai-agent-muse-home-address

404 Media — Humans Reading Copilot Prompts and Images https://www.404media.co/humans-reading-copilot-prompts-images/

Scott Wright’s Digital Legacy Network LinkedIn group https://www.linkedin.com/groups/39988004

** Watch this episode on YouTube **

https://youtu.be/rokXbwQgTgU

** Become a Shared Security Supporter **

Get exclusive access to bonus episodes, listen to new episodes before they are released, receive a monthly shout-out on the show, and get a discount code for 15% off merch at the Shared Security store. Become a supporter today by going to our YouTube channel's membership section: https://www.youtube.com/channel/UCg9CCDIYkDDqwEZ3UYaxjnA/join

** Thank you to our sponsors! **

Guardsquare

Special thanks to Guardsquare for sponsoring this episode! Guardsquare is the leader in mobile app

Listen Now

Love PodBriefly?

If you like Podbriefly.com, please consider donating to support the ongoing development.

Support Us