Meta’s Muse AI agent can run errands and make phone calls

In short: Meta has launched Muse in the US, a personal AI agent that can complete tasks across apps and even make phone calls, which raises questions about trust and privacy.

What happened

Meta introduced Muse in the United States on September 8, 2026. Meta says people can ask Muse to do tasks like sending email, booking travel, filling out forms, shopping, and making appointments. Muse is available in a Muse app, on the web, and in WhatsApp.

In a New York Times account, Muse did more than answer questions. It handled errands across services and by phone. The reporter said Muse called his dental insurer, gave his member ID, answered a security question, waited on hold, and then transferred him to a human representative.

The Times account and a separate Reuters report highlight a key feature, Muse can place calls to businesses. Meta has tested this for errands like booking haircuts, checking if a product is in stock, and asking contractors for price quotes.

Why it matters

This is different from a typical chatbot that only helps you write a message. Muse can act more like a personal assistant who logs in for you and speaks to other people on your behalf (like asking a helper to run errands using your wallet and your address book). To work well, it may need access to sensitive information such as email, and possibly banking or shopping accounts.

Meta says Muse runs inside Muse Secure VM, a dedicated virtual environment (a separate, locked room on a computer) and that it has safeguards. Still, these are company claims, not independent proof that all risks are removed. CNBC also reported that users must opt out if they do not want their Muse interactions used to train Meta’s AI models, so it is worth checking privacy settings before connecting accounts.

Source: NYTimes