Meta Pushes AI Agents Beyond Chat

Meta has launched Muse, an AI agent that can act across connected apps, moving the competition in artificial intelligence beyond chatbots and into software capable of completing tasks for users. The launch places greater emphasis on autonomy, platform reach and security as technology groups race to make AI agents part of everyday digital activity.

Muse can connect to email, calendars, payments, shopping, health and smart-home services, allowing it to send messages, book travel and complete transactions. It is initially available in the United States through a dedicated app and WhatsApp, with a free tier and paid subscriptions for heavier use.

For Meta, the product is also a test of whether its vast consumer ecosystem can become an advantage in the next phase of AI competition. The company is investing heavily in chips and infrastructure, and Muse offers a route to turn that spending into a service that reaches users across multiple applications rather than remaining confined to a single interface.

The wider industry is moving in the same direction. OpenAI, Anthropic and other developers are building agents designed to take actions, making access to apps and real-world services a new battleground. Meta’s scale across WhatsApp, Facebook and Instagram gives it reach, but that advantage depends on reliability and user trust.

Internal testing showed why the shift is difficult. Employees reported failures in task execution, repeated logouts and cases where sensitive data was exposed without authorisation. Meta delayed the release earlier this year to strengthen safeguards. The competitive question is now less about which company can build the most capable agent, and more about which can make autonomous AI useful without compromising privacy, security or control.