Meta launches AI agent that can access other apps to send emails, make payments

Meta has launched Muse, an AI agent capable of autonomously performing tasks such as sending emails, booking travel, and selling items on behalf of users, despite internal concerns over data security. The tool, central to CEO Mark Zuckerberg's vision of 'personal superintelligence,' is initially available in the US via a dedicated app or WhatsApp. It offers free and paid tiers at $20 and $100 per month. Internal tests revealed mixed results, including security flaws and reliability issues, while Meta's technical incidents have spiked 40% due to AI-related problems. The company aims to diversify revenue beyond advertising with this $130 billion AI investment.

Meta launched Muse, an AI agent that can autonomously send emails, book travel, and sell items on a user's behalf, despite internal concerns about data security. The tool, known internally as Hatch, is the centerpiece of CEO Mark Zuckerberg's plan to supply 'personal superintelligence' to Meta's billions of daily users. This is part of Meta's attempt to diversify revenue beyond advertising and build a business from its AI investments, which it forecasts will exceed $130 billion this year. Muse is available initially only in the US, via a dedicated app or WhatsApp. A basic version is free, with subscriptions at $20 and $100 per month. Users can opt out of training Meta's AI models with their interactions, and an encrypted version is planned. Muse is modeled on the open-source AI agent OpenClaw and runs on its own virtual machine. Meta built in a separate monitoring agent to prompt authorization before certain actions. Internal tests revealed mixed results: one employee found it useful for vacation logistics, while others flagged serious security flaws, such as an agent exposing iCloud photos after a prompt. Meta's CTO reported login issues, and another employee found the product unreliable for monitoring fast-selling items. Meta's internal technical incidents have spiked 40% from last year due to AI-driven coding and agent-related issues.