Meta is introducing new internal software that will record employees’ mouse movements, clicks, and keystrokes to support the training of its artificial intelligence models. The initiative, outlined in internal communications, is part of the company’s broader push to build AI systems capable of performing workplace tasks autonomously.
The tool, known as the Model Capability Initiative (MCI), will operate across selected work-related applications and websites, and will also capture occasional screenshots to provide contextual data. The goal is to improve how AI models replicate real human-computer interactions, such as navigating menus, using keyboard shortcuts, and completing routine digital tasks.
Meta stated that the collected data will be used exclusively for AI training and not for evaluating employee performance. A company spokesperson noted that such real-world interaction data is essential for building AI agents that can effectively assist users in everyday computing activities, adding that safeguards have been implemented to protect sensitive information.
The move is part of a larger internal transformation, where Meta is increasingly integrating AI into its workflows and restructuring roles around automation. The company envisions a future in which AI agents handle a significant portion of work tasks, with employees focusing on supervising and improving these systems.
However, the development has raised concerns around workplace surveillance and data privacy. Experts highlight that while such monitoring may be legally permissible in the United States with proper disclosure, similar practices could face regulatory challenges in regions with stricter data protection laws, such as Europe.
