Meta collects chat history, browser data, and employee clipboard content to train artificial intelligence

Meta ✴ collects employee work data to train AI

Internal Meta ✴ documents show that AI training logs include browsing history, clipboard actions, and employee correspondence across more than 200 apps. The goal is to teach AI to perform routine digital tasks autonomously. Success will largely depend on how regulators assess the difference between “behavioral data” and “personal information.”

What exactly does the MCI initiative collect?
* Model Capability Initiative (MCI) – Meta ✴’s program for developing AI models.
* Collects employee interaction data across more than 200 apps and online.
* Tracks mouse movements, clicks, navigation patterns – “workflow telemetry.”

This data is used to build AI agents that can replicate typical workflows and execute multi-step tasks in familiar software.

Why this raises concerns
1. Scope of collected data
* MCI records not only interface actions but also the content of emails and messages sent to employees in the U.S., even if they come from colleagues abroad.
2. Potential circumvention of international law
* If a U.S. employee uses GCath or e‑mail correspondence with someone outside the U.S., that activity is logged.
* Although the tool installs only on devices in the U.S. and is intended for behavioral analysis, it may unintentionally transmit international data into training.
3. Privacy
* Meta ✴ stated it notified non-U.S. citizens that the system runs on American colleagues’ computers and can capture their correspondence.
* The company claims compliance with personal data privacy laws but did not specify how incidental European messages are processed.
4. GDPR risks
* Even if data is collected only in the U.S., incidental collection of messages involving EU employees could trigger obligations under the General Data Protection Regulation (GDPR).

What employees say
* Some workers report that MCI captures a wide range of activity: code changes, browsing history, device sleep cycles, and clipboard actions.
* After installing MCI, there is a sharp increase in data usage.
* If this is true, Meta ✴ gains almost a complete view of how intelligent employees use tools—far more detailed than simple usage metrics.

How it fits into the company’s strategy
Meta ✴ increasingly focuses on automation.
MCI is part of large-scale efforts to create AI agents capable of taking over routine digital work: from navigating internal tools to performing repetitive tasks.

However, this approach already faces internal resistance. Some employees call the initiative an aggressive attempt to turn human workflows into machine-readable systems.

Bottom line: Meta ✴ collects extensive employee work data to train AI, raising privacy and international regulation concerns. The program’s success will hinge on how regulators assess the boundary between behavioral data and personal information.