More than Big Brother, Meta will record employees’ keystrokes and mouse activity to train AI

The relentless pursuit of data to train AI models pushes big tech companies toward new and controversial frontiers.

Meta has decided to look within its own walls, transforming the daily activities of its employees into an extensive source of information.

The U.S. tech giant, owner of WhatsApp, Facebook and Instagram, has begun installing a tracking software on employees’ computers in the United States in order to record the mouse movements, clicks and individual keystrokes.

Meta tracks US employees’ mouse and keyboard activity to train AI

facebook passkey
Credits: Meta

The new tool, named Model Capability Initiative, will operate continuously on applications and websites related to work activity, even reaching to acquire periodic screenshots of the monitors.

The rationale behind this decision is technical: current models still struggle to naturally replicate human interactions with graphical interfaces, such as navigating dropdown menus or the smooth use of keyboard shortcuts.

Through internal communications, company executives clarified that this meticulous tracking will enable the creation of software capable of operating with great autonomy. Andrew Bosworth, Meta’s Chief Technology Officer, outlined a future scenario in which machines will handle most operational tasks.

In this vision, human staff will take on a directive and supervisory role, intervening only when necessary to correct and refine the actions of the programs, so that they learn from their mistakes.

Between corporate assurances and staff layoffs

In light of the inevitable privacy concerns, Meta’s spokespersons stressed that the data collected will in no way be used to assess employees’ job performance.

The company also assured the presence of strict filters to protect sensitive content, although it did not provide technical details on which categories of information will enjoy this protection during screen capture.

This massive push toward automation sits within a phase of strong occupational contraction for the tech sector. Numerous entities are significantly downsizing their staff, replacing workers with automated systems.

Meta has planned a 10% cut to its global workforce, following the example of other U.S. companies like Amazon and Block, which recently reduced tens of thousands of their employees.

Concurrently, in Menlo Park a more generic professional role is taking shape, defined as “AI builder“, designed to delegate writing of programming code to the new generative models in testing phase.

The legal issue and workplace surveillance

Historically, activity logging systems were used by cybersecurity divisions to prevent illicit behavior, but current monitoring brings surveillance to a decidedly more pervasive level. In the United States, federal law grants companies broad latitude, imposing no stringent limits on worker monitoring.

The situation is radically different in Europe, where the initiative would clash with legal obstacles almost insurmountable. The General Data Protection Regulation (GDPR) would severely limit data collection of this nature.

In countries like Italy, the use of electronic tools to track employee productivity or daily activity is regulated in a very restrictive way, while in countries like Germany the use of keystroke loggers is tolerated exclusively in the presence of well-founded suspicions of serious criminal offenses.