Meta will monitor employees' keystrokes and mouse clicks to train its AI models

Meta has built an internal tool that captures employee keyboard and mouse activity—converting these inputs into training data for its artificial intelligence systems. This marks a shift toward using real workplace interactions as a data source for AI development.

Unlike traditional AI training that relies on public datasets or labeled examples, Meta's approach harvests continuous behavioral data from actual work—meaning the AI learns from patterns in how people actually type, click, and navigate tasks. This can make AI assistants better at mimicking human workflow and decision-making.

Employees at Meta and other companies using similar tools should know their keystroke patterns are now being collected and used to improve AI products. If your workplace adopts comparable monitoring, your typing speed, corrections, search queries, and clicking habits become part of a company's AI training pipeline—raising questions about what behavioral data gets captured and how long it's retained.