As Meta asks people to trust its AI with more of their everyday lives, Mark Zuckerberg used his Connect 2026 keynote to outline new protections for the information those products collect.
For Muse, Meta’s personal AI agent, that includes an upcoming system called Muse Confidential VM. Each Muse already runs on a dedicated computer in the cloud, known as Muse Secure VM. The planned confidential version adds encryption designed to prevent even Meta from accessing the user’s data.
The distinction matters for an assistant designed to work across connected accounts and handle tasks such as sending emails, booking travel, and shopping.
Mashable Light Speed
Zuckerberg also addressed privacy on Meta’s smart glasses, pointing to safeguards he said were built into the devices from the beginning. Those include an indicator light that turns on when someone captures a photo or video, and a feature that disables video recording if the light is tampered with.
Meta is now developing private processing for certain AI features on the glasses. Zuckerberg described a protected server environment where personal data would be inaccessible to others, including Meta, using technology also employed for WhatsApp.
He said the feature would roll out soon, starting with private translations.
The announcements address a growing question for Meta: how to offer increasingly personal AI features without asking users to accept increasingly broad access to their information. For glasses, there’s another audience to convince, too — the people on the other side of the camera.