On Tuesday, Meta launched Muse in the US as a personal AI agent for adults, pitching it as software that can turn schedules, shopping and long-term ambitions into completed tasks. CEO Mark Zuckerberg called it “the personal agent that understands your goals and works 24/7 to get things done for you.”Mark Zuckerberg (finkd)¹

Muse is meant to move beyond the familiar chatbot: users can hand it a goal, then let it browse, fill forms, send emails, plan travel, negotiate and return for approval when money or other sensitive decisions are involved. Meta says the free product will be available on iOS, Android and the web, with $20 and $100 monthly tiers for heavier users.² Zuckerberg framed the rollout as an opening move toward “personal superintelligence,” offering a free allowance of up to 100 million tokens a week.Mark Zuckerberg (finkd)³

The company’s sales pitch rests on control. Muse runs on a dedicated cloud virtual machine with a visible browser, while a separate system, Sentinel, determines whether proposed actions are allowed, blocked or require the user’s sign-off. Meta says people can narrowly limit permissions and opt out of having queries used to train its models; it is also promising a confidential workspace that even Meta cannot inspect later this year.²

But the design exposes the central dilemma. For Muse to be genuinely useful, it may need access to email, calendars, payment tools and the habits that make up a private life. Reporting on the launch notes that Meta is asking for “significantly more trust than social media ever did,” after privacy failures ranging from the Cambridge Analytica scandal to FTC actions and a recent multistate settlement over consumer harms.⁴

Meta is betting that granular permissions, separated credentials and useful automation can change that calculation. Consumers, however, will decide whether an agent built to remember more, act longer and connect deeper is a breakthrough—or simply a larger version of an old trust problem.