Meta is taking a more direct route in building its AI systems, and this time it starts from inside the company.
Meta is reportedly introducing tools that will track employees’ screens and keystrokes as part of efforts to train and improve its artificial intelligence systems.
The approach is designed to capture how work actually gets done, from typing patterns to on-screen activity, giving AI models more real-world data to learn from.
This marks a shift from relying mainly on external datasets to using internal workflows as a training ground.
For Meta, the logic is straightforward.
If AI is expected to assist with real tasks, it needs to understand how those tasks happen in practice, not just in theory.
But the move raises immediate concerns.
Monitoring employee activity at this level brings up questions around privacy, consent, and how far companies should go in collecting behavioural data, even within their own workforce.
It also highlights a growing trend in the AI race.
Companies are no longer just competing on models, they are competing on data. And increasingly, that data is being sourced from environments closest to them.
For employees, it introduces a new layer to the workplace, where everyday actions could double as training input for systems being built around them.
The balance now sits between innovation and oversight, and how companies manage that line will likely shape how these systems are accepted internally and beyond.
