Meta has expanded data collection from employees' work devices in the US as part of an initiative to train artificial intelligence (AI) systems. Internal documents show that the tool captures not only user actions within programmes, but also other types of activity, including correspondence and browsing history.
WHAT IS KNOWN ABOUT THE PROJECT
According to Reuters, internal Meta materials indicate that the Model Capability Initiative (MCI) collects data on employee interactions with more than 200 apps and websites. The project aims to create AI systems capable of independently performing standard digital tasks.
At a basic level, the system tracks cursor movements, clicks, and navigation within programmes. As the documents show, over time such data could be used to train AI to perform multi-step operations in work applications.
WHAT DATA HAS COME UNDER SCRUTINY
The system can also capture the content of emails and messages sent to employees based in the US, including correspondence with colleagues from other countries. One of the company's internal FAQs noted that if a US-based employee uses the tool while communicating via email or chat with an overseas colleague, such activity may be recorded.
Meta spokesperson Dave Arnold stated that the tool is only installed on devices in the US and is primarily intended for analysing user behaviour, not the content of communications. According to him, the company has assessed the privacy risks and intends to comply with applicable laws and regulatory requirements.
WHAT QUESTIONS HAVE ARISEN WITHIN THE COMPANY
Some employees reported that the system operates as a larger-scale data collection mechanism embedded within existing monitoring software. Internal analyses referenced in the documents indicate that MCI may capture code changes, browser history, device sleep/wake cycles, and clipboard activity.
One employee described the project as a system capable not only of performing individual actions but also of understanding the sequence of operations when working with documents and programmes. The publication containing this description was later deleted, according to the materials.
Dave Arnold called such assessments "fundamentally inaccurate", but did not comment on specific technical details.
WHY THE PROJECT HAS ATTRACTED THE ATTENTION OF DATA PROTECTION EXPERTS
Some experts link the situation to the requirements of European data protection legislation. Even if the project's infrastructure is limited to US territory, correspondence with employees from the European Union could implicate the norms of the EU General Data Protection Regulation (GDPR).
Kleanthi Sardeli, an expert at the organisation NOYB, stated that using work communications to train AI may not align with the original purposes for which such data was collected. In response, Meta told the Irish data protection regulator that collecting information on employees from the EU is not a primary objective of the project.