Meta Uses Its Employees as Guinea Pigs for Its AI

Meta Uses Its Employees as Guinea Pigs for Its AI

Meta, the parent company of Facebook, is developing an ambitious project aimed at enhancing its artificial intelligence (AI) agents by analyzing its employees' interactions with their computers. According to internal notes obtained by Reuters, the company is using tracking software to record employees' mouse movements, clicks, and keystrokes.

This software is integrated into various internal professional tools and regularly captures screenshots of the applications and websites used for work. Meta's goal is to place each action in context to better understand how humans interact with machines.

Meta's Justification for This Data Collection

A spokesperson for Meta explained that this data is essential for improving AI agents' ability to perform concrete tasks. Training AI systems requires real-world examples to navigate menus, click buttons, and replicate simple yet crucial actions.

Meta emphasizes that this data is not used to evaluate employee performance, but the initiative remains a continuous observation that does not have unanimous support. Some employees may feel uncomfortable with the idea of becoming direct sources of training data for AI.

A Collective Contribution or Exploitation?

Meta presents this project as a form of collective contribution, highlighting that texts, images, and videos available online are already widely used for training AI models. However, real interaction data with a computer is rare and valuable.

Other companies simulate complete environments to generate this behavioral data, but Meta has chosen to leverage the actual interactions of its employees. Furthermore, the company is reportedly considering setting AI usage targets for certain employees and may reduce its global workforce by 10%, according to some rumors.