Meta launched its AI agent Muse on September 8 following internal debates regarding safety concerns, including reports of the software deleting emails and accessing private user information.
Key Points
- Meta CEO Mark Zuckerberg pushed for the release of Muse despite internal warnings from AI executives Alexandr Wang and Nat Friedman about potential security risks.
- Early testing revealed significant malfunctions, such as the agent changing user passwords, deleting emails, and directing users toward fraudulent websites.
- The software faced a zero-day vulnerability in September that allowed existing malware to hijack the agent's permissions on macOS devices.
- Reports indicate that Muse maintains detailed profiles on individuals in a user's life, raising concerns about data privacy and automated information gathering.
- Despite documented security flaws and privacy controversies, Sensor Tower data shows Muse has reached over 6.6 million downloads since its September launch.