Meta Platforms is reportedly tracking employee keystrokes, mouse movements and on-screen activity through aninternal artificial intelligence training programmedesigned to improve its next generation of AI systems, according to multiple reports.
The initiative, which applies to US-based employees using company-issued devices, has drawn scrutiny over workplace surveillance and how far tech firms are going to train AI agents.
Reports published this week suggest the system captures general digital behaviour across work environments, rather than targeting specific websites such as Google, LinkedIn or Wikipedia.
The initiative, internally referred to as the Model Capability Initiative (MCI), is designed togather real-world interaction datafrom employees as they work. The system records keystrokes, mouse movements, click patterns and occasional screen snapshots while staff use work-related tools and web-based platforms.
According toCNBC, Meta's internal AI training initiative involves tracking employee activity across commonly used platforms such as Google, LinkedIn and Wikipedia as part of efforts to improve its artificial intelligence systems.
The data is collected from company-issued laptops and is intended to help train AI systems to understand how humans interact with digital interfaces in real working conditions. This includes how users navigate software, switch between tasks and engage with productivity tools.
Meta's objective is to improve AI agents capable of performing complex workplace functions, such as automating administrative processes and interacting with enterprise software in a human-like manner.
Online discussions have linked the monitoring programme to specific websites including Google, LinkedIn and Wikipedia. However, current reporting does not confirm that these platforms are being individually targeted or singled out.
Instead, these sites are frequently used as examples of common workplace browsing activity. Employees may access search engines, professional networking platforms or reference sites during daily tasks, and such interactions could be indirectly captured if they occur on monitored work devices.
The system is not understood to operate as a website blacklist or targeted surveillance tool. Rather, it records broader behavioural signals such as typing and navigation patterns across digital environments used for work.
Source: International Business Times UK
