Meta’s Muse aims to regain consumer trust amid aggressive AI push

Meta just pulled off something rare in Big Tech: it launched an AI product that people actually want to use. Muse, the company’s personal AI agent, debuted on September 8 and immediately climbed to the top of the US iOS free app charts. The timing is conspicuous. Meta is fresh off an $18 billion multistate settlement over social media harms to youth, and Muse arrives wrapped in privacy-first messaging. What Muse actually does Unlike the wave of chatbots that emerged after ChatGPT’s debut, Muse is designed to do things rather than just say things. It sends emails, books travel, manages purchases, and integrates with services like Gmail, Spotify, and Stripe. The underlying architecture runs on what Meta calls Muse Spark models. Each agent session operates inside a Muse Secure VM, essentially a sandboxed virtual machine that gives the AI browser access for tasks that don’t have a direct API. When the agent needs to do something sensitive, like entering payment information or accessing personal data, it requests explicit user permission first. The pricing structure follows the now-standard AI tiering playbook. A free version handles basic tasks, a $20 per month Power tier unlocks expan...