Meta launches Muse personal AI agent

Meta has launched Muse, a consumer-focused personal AI agent designed to offload routine digital work such as booking appointments, completing electronic forms and monitoring home security camera feeds. The company positions Muse as an assistant that can handle complex workflows behind the scenes while presenting a simple, approachable interface to users.

The agent is powered by the Muse Spark family of foundation models, which underpin the app’s ability to coordinate tasks across different services and data sources. Meta developed Muse to perform what it describes as heavy lifting in coding and integrations while keeping the experience straightforward for non-technical users.

Muse is being offered in multiple tiers, including a free option and subscription plans priced at $20 and $100 per month, with charges tied to usage levels. This structure is intended to accommodate light users alongside those who require higher volumes of automated assistance.

The personal agent is available to consumers in the United States through native apps on iOS and Android as well as a standalone website. Muse is also accessible via WhatsApp, giving users the ability to interact with the agent in a familiar messaging environment.

While WhatsApp provides the core conversational functionality of Muse, it does not include certain features available in the dedicated app. Those app-only features include a personalized feed and an ideas tool that surfaces example tasks and tips for using the agent.

Within the Muse app, users can view a feed that aggregates and summarizes updates from their connected Facebook and Instagram accounts. The feed can also pull in selected articles or stories from the web, depending on the preferences the user sets for what should be included and summarized.

Beyond passive updates, Muse is designed to act on behalf of users by carrying out digital tasks across connected services. These capabilities include actions like submitting information through online forms or keeping an eye on smart-home camera feeds, with the goal of centralizing task management within a single AI agent.

Meta emphasizes that each Muse personal agent runs within its own isolated environment inside the company’s computing infrastructure. The company states that the agent does not see actual passwords or payment details and that it prompts users before performing sensitive actions.

Users are given the option to opt out of having their interactions with Muse used to train Meta’s AI models. If users do not opt out, Meta says it will first remove critical personally identifying information from conversations and interactions before using them to improve its models.

To strengthen security, Meta is expanding Muse into the scope of its public bug-bounty program. Third-party security researchers are being encouraged to search for vulnerabilities in the personal agent, with financial rewards offered for qualifying findings.