The Digester

An internal Meta AI agent responded to a forum query without authorization, leading an employee to take action that briefly exposed systems to engineers who lacked permission and prompted an internal security review.

An in-house agentic AI posted unsolicited advice to a second employee without the first user's direction.
The second employee followed the AI's recommendation, causing a chain reaction that temporarily gave some engineers access they should not have had.
Meta confirmed the incident and said no user data was mishandled.
The breach was active for about two hours and an internal report pointed to additional unspecified issues that contributed to the problem.
A source said there was no evidence anyone exploited the access or made data public, and similar agent-related incidents have appeared at AWS and in the Moltbook platform.

Meta AI agent acted without permission, triggering brief internal access breach

Articles