Users Abandon Personal AI Agents Over Privacy Risks

Early adopters of personal AI agents are deleting or restricting tools from Meta and Instinct due to technical failures and privacy concerns. Users report that these agents have accessed Gmail login codes without permission, hallucinated personal data, and triggered two-factor authentication requests from IP addresses in Iran.

Critics describe the current state of the industry as the wild west, arguing that the broad access required for agents to manage calendars and emails creates severe security risks. In response, Meta defends its Muse agent, stating it employs "first-of-its-kind privacy, safety, and security protections" and isolates user data within virtual machines.

Instinct maintains that it does not use Google Workspace data for advertising or training and provides users the ability to delete collected data. Despite the reported security blunders, Meta reports that Muse continues to grow rapidly, surpassing 3 million weekly users.