Meta launched Muse on Tuesday as a personal artificial-intelligence agent for U.S. users age 18 and older. The company says the service can help with schedules, shopping and longer projects by breaking a goal into a plan and advancing tasks over time. It is available through a separate Muse application and through WhatsApp messages.

Meta described functions ranging from sending an email and booking travel to creating a yearlong exercise plan or organizing steps for a new business. Those examples are company representations of intended capability, not an independent measurement of completion rates or error frequency. Availability may also vary by account, service connection and the permissions a user grants.

The company says each user’s agent and data operate inside a dedicated secure virtual machine. Meta is emphasizing privacy and safety controls, but the launch reporting did not provide a third-party audit, breach history or comparative test of the isolation design. Security claims therefore remain attributed to Meta pending independent evaluation.

Agent systems differ from ordinary chatbots because they can maintain goals, use connected services and take actions after the initial prompt. That creates more operational state: credentials, messages, reservations, schedules and intermediate decisions. A mistaken answer can be corrected in text, while an incorrect external action may require cancellation or recovery.

The launch establishes product availability and Meta’s stated architecture. It does not yet establish how Muse performs on adversarial instructions, ambiguous requests, long-running plans or service failures. Usage reports, technical documentation and independent testing will be needed to evaluate reliability and the practical strength of the safeguards.

The checked record for Meta Launches Muse, an AI Agent Designed to Act Across Long-Term Tasks establishes the following dated facts: Muse launched Tuesday for U.S. users age 18 and older. It can be contacted through a separate app or WhatsApp. Meta says it can send email, book travel and coordinate long-term plans. Meta says each account uses a dedicated secure virtual machine. The launch is initially limited to the United States. The relevant background is also specific: Agent systems can take actions through connected services rather than only answer questions. External actions can require cancellation or recovery when mistaken. Independent launch-day testing was limited. Privacy, security and reliability claims were largely supplied by Meta and lacked broad independent verification at launch. The cited reporting does not resolve questions beyond those stated limits.