Meta Requires US Employees to Install AI Training Tool That Logs Keystrokes, Screenshots on Work Computers

Meta has begun installing a workplace monitoring tool on US-based employee computers that captures detailed behavioral data—keystrokes, mouse movements, clicks, and periodic screenshots—explicitly to train the company's AI agents. The initiative, called the Model Capability Initiative (MCI), represents a direct conversion of human labor patterns into training material for automated systems.

The tool operates continuously within work-related applications and websites, recording how employees interact with software interfaces. Meta intends to use this data to train AI models that can perform computer-based work tasks autonomously, essentially reverse-engineering human workflow into machine capability. According to the reporting, Meta has stated the collected data will not be used for employee performance evaluations—though the distinction between behavioral monitoring for AI training and traditional performance surveillance raises questions about implementation boundaries and employee consent.

The move places Meta among technology companies aggressively pursuing AI systems capable of operating software the way human workers do. By using its own workforce as a training ground, the company gains real-world operational data at scale, bypassing the need to contract external data-labeling services. The initiative signals growing industry interest in capturing observable human-computer interaction patterns as a resource for developing autonomous agents, while simultaneously exposing workers to persistent behavioral surveillance under corporate mandate. Legal and privacy implications remain under scrutiny as the practice expands.