Meta will monitor employees' mouse and keyboard activity to train AI agents, exposing the scarcity of real-world interaction data for AI development

Meta is planning to track employees' computer activity—mouse clicks, keyboard inputs, and screen interactions—to collect training data for AI agents (AI systems that can take actions on their own). This shift highlights that AI developers struggle to find enough real, diverse examples of how people actually work.

Rather than asking employees to manually label data or simulate tasks, Meta will capture authentic workplace behavior at scale. This means the company gets messy, unscripted examples of real problem-solving—not clean, staged scenarios—which typically produces more robust AI agents that handle unexpected situations better.

If this approach works, companies building AI assistants could deploy smarter tools across offices faster, reducing the months-long data collection projects that currently slow down AI product launches. For employees, it signals that workplace monitoring may become standard practice as companies harvest behavioral data to train internal AI systems.