Meta investigates leak of employee monitoring data affecting 45,000 records

Meta is investigating an incident in which data collected from employee computers may have been accessed by multiple internal staff members. The breach involves approximately 45,000 data tables that could contain keystrokes, mouse clicks, screen content, AI prompts, private chats and performance‑evaluation information. The data was gathered under a program launched in April to train AI models to use software in a human‑like way, and it applies to Meta employees working on computers in the United States.

Spokesperson Tracy Clayton confirmed the investigation and said the program was designed with privacy safeguards, though no evidence of unauthorized access has been found so far. The issue follows a petition signed by more than 1,600 staff members warning that such monitoring poses security, legal and privacy risks. CTO Andrew Bosworth acknowledged that the rollout did not meet Meta’s privacy standards and pledged to share the investigation results with employees. The episode adds to internal tensions as Meta continues workforce reductions and a shift toward AI development.