Meta pauses employee‑monitoring AI program after sensitive data leak

Meta halted its internal Model Capability Initiative (MCI), a program that recorded employees’ mouse movements, keystrokes, screen captures and other on‑screen activity to train artificial‑intelligence models. The pause follows a security incident in which the collected data—including private conversations, performance metrics and personal information—was inadvertently exposed to other staff members.

A Meta spokesperson said the company had designed the program with privacy safeguards but, “we have no indication that any employee has accessed the data improperly, and we are pausing it while we investigate,” citing the internal security report. The initiative, launched in April for U.S. staff, was part of a broader effort to develop AI that could eventually automate office tasks. The incident has raised concerns among employees about workplace surveillance and data privacy.

Meta confirmed that it has fixed the access flaw but has not disclosed whether the leaked information was misused. The company faces scrutiny as it seeks to accelerate AI development while managing internal privacy risks.