Meta introduces privacy features for AI assistant Muse

Meta just did something it has historically struggled to do: make a convincing privacy pitch. The company launched Muse, a personal AI agent, on September 8, 2026, and the central selling point isn’t what the agent can do. It’s what Meta promises not to do with your data. The market response suggests investors are buying the pitch. Meta’s stock surged more than 20% following the announcement, and Muse racked up roughly 2.8 million installs in its first 12 days. What Muse actually does Muse isn’t another chatbot that answers trivia questions and writes mediocre poetry. It’s an AI agent, meaning it acts on your behalf across connected applications and a built-in browser. Send emails, shop online, book services. The subscription model comes in three tiers: free, $20 per month, and $100 per month, with different levels of capability at each price point. The initial rollout covers users aged 18 and older in the US, accessible through iOS, Android, web interfaces, and WhatsApp integration. At Meta Connect 2026, held September 23-24, the company revealed partnerships with Walmart and PayPal to expand what Muse can actually transact on. The privacy architecture Each user gets their own iso...