Meta Platforms is pushing deeper into its AI-first strategy by introducing a new internal system that tracks how employees use their computers, including mouse movements, keystrokes and screen activity, to train its artificial intelligence models.
The initiative is part of a broader effort to build AI systems that can replicate real human interactions with software. By observing how employees navigate tools, use shortcuts and complete everyday tasks, Meta is trying to generate more accurate training data for AI agents designed to eventually perform similar work autonomously.
The tracking will be enabled through internal software deployed on work devices, focusing on activity within job-related applications. In some cases, the system may also capture snapshots of on-screen activity to better understand context and decision-making patterns. The goal is to address gaps where AI still struggles, particularly in handling real-world workflows and interface navigation.
This move comes as Meta continues to reorganise itself around AI, with a growing emphasis on automation and efficiency. The company is working towards building AI agents that can take on routine tasks, potentially changing the nature of many roles internally.
However, the rollout has also raised concerns among employees and industry observers. Questions around workplace surveillance, data privacy and consent have come into focus, especially given the scale and granularity of the data being collected. While Meta has indicated that safeguards are in place and that the data will be used for training rather than performance evaluation, the lack of opt-out options for some employees has added to the discomfort.
The development highlights a larger shift across the tech industry, where companies are increasingly relying on real user behaviour to improve AI systems. At the same time, it brings into focus the trade-off between accelerating AI capabilities and maintaining clear boundaries around employee monitoring in the workplace.
