Meta adds email addresses and video chat to Muse AI agent

By Splitfeed EditorialLast updated Sep 26, 2026, 19:17 UTC

On September 23, 2026, Meta announced a set of updates to Muse, its AI agent product. The additions include dedicated email addresses for each Muse agent, which the agent can use both to carry out tasks and to communicate directly with users. Meta is also adding video chat as an interaction mode, alongside what The Decoder described as video avatars. Separately, the Muse Mac app, which had launched recently, is being extended to allow the agent to control the user's computer, giving it direct access to the operating system rather than limiting it to a conversational interface.

The combination of these features marks a meaningful shift in what kind of entity Muse is positioned to be. An AI agent with its own email address can operate in contexts where the user is not present, sending and receiving messages on their behalf. Computer control on Mac pushes further in the same direction, allowing Muse to take actions inside applications rather than simply responding to prompts. Video chat adds a real-time social dimension. Together, the updates move Muse closer to what researchers and product teams call an agentic model, one that acts in the world rather than sitting inside a chat window. The open question is how much autonomy users are actually granting, what oversight exists over actions taken via email or on the desktop, and how this compares to similar agentic products from competitors.

Because no reactions have been published yet from the Pro-AI, Anti-AI, or Middle Ground camps, their likely positions can only be anticipated. Proponents of AI development would typically welcome the expanded capabilities as practical productivity gains, treating autonomous email and computer control as a natural evolution of the assistant model. Critics would likely raise concerns about an AI agent acting independently in email and on a personal computer, focusing on privacy, the risk of unintended actions, and the difficulty of auditing what the agent does. A middle-ground perspective would probably acknowledge the usefulness while calling for clear user controls, transparency about what the agent can access, and meaningful opt-in mechanisms before any autonomous action is taken.

The details most worth watching are how Meta implements permission controls for computer access and email, whether any regulatory or platform-level scrutiny follows, and how user uptake of the Mac app's control features develops in the weeks after the September 23 announcement.